Sie sind auf Seite 1von 8

Limiting Oxygen Concentrations of Gases

Laurence G. Britton,a Martin P. Clouthier,b,c Benjamin K. Harrison,d and Samuel A. Rodgerse


a
AIChE Fellow, Charleston, WV; lbritton.consult@suddenlink.net (for correspondence)
b
Fauske & Associates, Halifax, NS, Canada
c
Department of Process Engineering and Applied Science, Dalhousie University, Halifax, NS, Canada
d
Department of Chemical and Biomolecular Engineering, University of South Alabama, Mobile, AL
e
Honeywell, Colonial Heights, VA
Published online in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/prs.11806

This article discusses technical and practical aspects of Keywords: limiting oxygen concentration; limiting oxygen
the limiting oxygen concentration (LOC) including a concentration measurement; flammability; inerting
detailed appraisal of possible sources of measurement error.
LOC values obtained in the “classical” 5-cm diameter, verti-
INTRODUCTION
cal tube apparatus have been described as nonconservative
Chapter 7 of NFPA 69 “Standard on Explosion Prevention
owing to flame quenching. However, much of the difference
Systems” (2014) presents formal procedures for “Deflagration
between classical LOC values and those obtained in closed Prevention by Oxidant Concentration Reduction.” The gen-
vessels using ASTM E2079 is due to the change in the defini- eral practice is usually known as “inerting,” which involves
tion of “ignition.” Formerly defined as propagation of a visi- the addition of inert gas to a mixture of fuel and oxygen so
ble flame through a vertical tube at least 1.5 m tall, which that the oxygen stays some specified margin below the
in a closed tube would typically generate an overpressure in “limiting oxygen concentration” or LOC. The LOC is defined
excess of 100%, ignition is now defined as an overpressure as the smallest concentration of oxygen that can support a
that exceeds 7%. The small pressure rise criterion tends to self-propagating flame whatever the concentrations of fuel
underestimate the LOC in smaller test vessels, especially for and diluent. This article focuses on oxygen and nitrogen
tests made at elevated pressures and temperatures. More use although inerting can be applied to other oxidants and inert
should be made of ASTM E2079’s flexibility regarding the use gases. Nitrogen is the most commonly used inert gas, as
of larger pressure rise criteria. Having lost sight of the flame, described, for example, by Yanisko et al. [1].
it is important not to lose sight of the objective, namely to Annex C of NFPA 69 contains “additional information” on
select test criteria that correctly predict the flammable limits inerting and includes a widely used table of LOC values for
pertaining to large volumes of gas mixture. Additional gaseous fuels [2]. In the 2008 edition, most of the tabulated
“reference quality” LOC data are needed both to improve the LOC values were decreased by 2 mol % oxygen, regardless
definition of “ignition” with respect to test vessel volume and of their initial magnitude. This was prompted by LOC data
to investigate the accuracy of LOC estimation techniques. recently measured in a large (120-L) test vessel using the cur-
Any large-scale reference quality test program should investi- rent standard test method ASTM E2079-07 [3] and considered
gate the effect of low levels of forced turbulence on measured state-of-the-art. New LOC data for hydrogen, carbon monox-
LOC values. ide, methane, and propane [4] were 7.5 6 0.5% smaller than
The NFPA “Explosion Protection Systems” Technical Com- “classical” values measured using the 5-cm diameter, vertical
mittee recently supported the development of two Tentative tube apparatus. However, the new LOC for ethylene was
Interim Amendments (TIAs) to NFPA 69 “Standard on Explo- 15% smaller (8.5 mol % oxygen as opposed to the classical
sion Prevention Systems.” The first TIA resolves issues with 10.0 mol % value). It was hypothesized that the 5-cm vertical
the table of LOC values for gases, which had been modified tube may have caused excessive quenching at the LOC and
and over-corrected in the 2008–2014 editions. The second to offset this effect, the NFPA 69 Technical Committee
TIA introduces a simple method for estimating the LOCs of adjusted all classical LOC values downward by 2 mol % oxy-
single gases and gas mixtures based on the equation gen. LOC values measured using ASTM E2079-07 or appara-
LOC 5 LFL 3 S, where LFL is the lower flammable limit and S tus approximating to its requirements were not adjusted.
the molar stoichiometric ratio of oxygen to fuel. An Appendix Surprisingly, there were no public comments about these
describes the two TIA recommendations while a second changes although not all Technical Committee members
Appendix describes how the estimating methods have been agreed with the approach. There’s no direct evidence that
incorporated into ASTM’s “CHETAH” Program. V C 2016 Ameri- the 5-cm vertical tube causes more flame quenching at the
can Institute of Chemical Engineers Process Saf Prog 35: 107–114, LOC than it does at the lower flammable limit (LFL) and
2016 upper flammable limit (UFL), the likelihood of which was
discounted by the U.S. Bureau of Mines following extensive
Additional Information in the form of Appendices A and B may be
testing in larger diameter tubes; see, for example, Lewis and
found in the online version of this article. von Elbe [5].
This article makes the case that much of the LOC discrep-
C 2016 American Institute of Chemical Engineers
V ancy can be attributed to a difference in LOC definition, since

Process Safety Progress (Vol.35, No.1) March 2016 107


the classical 5-cm vertical tube test method required a flame to
propagate the full length of a tube at least 1.5 m tall (which in
a closed tube would typically produce a pressure rise in excess
of 100%) while the ASTM E2079 closed vessel test method
requires only a 7% pressure rise. Indeed, in cases where Zlo-
chower and Green observed a pressure rise of about 30% the
corresponding LOC values (apart from ethylene) were much
the same as the classical ones. Moreover, it’s proposed that the
classical LOC data were subject to a number of experimental
errors that were deliberately avoided by Zlochower and Green.
In all cases where experimental errors resulted in overesti-
mated classical LOC values, the experimental error and the
“definition” error become additive. This is because the new
definition always results in smaller LOC values.
Ethylene’s classical LOC value of 10.0 mol % oxygen
derives from two studies reviewed by Coward and Jones [6];
the first involved mixtures of ethylene, oxygen, and nitrogen
while the second involved mixtures of ethylene, air, and nitro-
gen. In both cases, a 5-cm vertical tube apparatus was used.
In the first case, a fuel-oxygen flammability diagram was pre-
sented that directly gave the LOC as 10.0% oxygen. In the sec-
ond case, a flammability diagram of fuel versus added
nitrogen was presented, showing the LOC was reached after
the addition of 50% nitrogen. No grid was shown but the eth-
ylene concentration at the LOC appears to be about 3.5%.
The LOC is therefore approximately 9.7% oxygen, which is
roughly the same value found in the other study. Since a 5-
cm vertical tube should not quench near-LOC ethylene flames
to a greater extent than CO, H2, CH4, and C3H8, we propose
instead that both classical studies were flawed; the fact that
different workers found roughly the same LOC value is con-
sidered to be a coincidence rather than a confirmation. We’ll
explore sources of error later in this article but one reason
why two similarly high LOC values should have been
obtained for ethylene is that two similarly weak ignition sour-
ces might have been used. The most frequently used ignition
source was a continuous arc as opposed to a discrete spark
having a high power density; it’s plausible that “weak
ignition” errors are more pronounced at the LOC than at the
more commonly measured LFL. Another possible reason is
that unlike most hydrocarbons, ethylene’s LOC occurs at a Figure 1. Schematic Cartesian flammability diagrams for (a)
fuel concentration not far above the LFL; the so-called “nose” propane and (b) carbon monoxide in air plus diluent
region shown in typical flammability diagrams such as those (Adapted from Ref. 6, with permission from U.S. Bureau of
of Kuchta [7] is relatively elongated and narrow. Conse- Mines). [Color figure can be viewed in the online issue,
quently, an insufficient number of tests might have been per- which is available at wileyonlinelibrary.com.]
formed near the optimum LOC composition.
Following publication of the 2014 edition of NFPA 69, we
proposed two Tentative Interim Amendments (TIAs) that LOC in Relation to Flammability Diagrams
were favorably received by the Technical Committee. The Let’s first consider the LOC in relation to a flammability
first TIA changes the LOC adjustment process and the associ- diagram. The LFL is the smallest concentration of fuel that
ated table of LOC values; this effectively increases all of the can propagate a flame when mixed with an oxidant and is
classical LOC values that were previously overcorrected. The found to be approximately the same for both oxygen and
second TIA describes a technique for estimating LOC values air. This is because the oxidant (oxygen) and diluent (nitro-
from LFL values, plus worked examples showing how to gen) have almost identical heat capacities and in both cases
estimate LOCs of gas mixtures starting either from measured there’s an overabundance of oxygen, since the oxygen con-
LOC or LFL values. The detailed TIAs are presented in centration greatly exceeds the “stoichiometric” concentration
Appendix A. The LFL and LOC of single fuels can be calcu- needed for complete combustion. As additional nitrogen is
lated for many fuels using ASTM’s “CHETAH” Program with- added to the mixture, there’s negligible increase in the LFL
out reference to test data. Appendix B describes how until the oxygen concentration is decreased sufficiently to
CHETAH has been updated to allow similar estimates to be become rate controlling. Beyond this point, the lower limit
made for mixtures. concentration of fuel increases above the LFL and the flam-
mable region forms the familiar “nose” of a flammability dia-
gram. This is illustrated for propane by Figure 1a. At the
DISCUSSION LOC, the LFL curve joins the UFL curve as shown in Figures
We will now discuss technical aspects of the LOC concept 1 and 2. As illustrated by Figure 1b for carbon monoxide,
and practical application, including a critical review of cur- fuels whose diffusivities are closer to that of oxygen exhibit
rent LOC measuring techniques. LOC values closer to the point P. As discussed later, this

108 March 2016 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.35, No.1)
Process Applications of LOC and MOC
The LOC was formerly known as the minimum oxygen con-
centration (MOC). However, to avoid confusion the term “MOC”
should be reserved for certain process applications where fuel
concentrations are held at or above some minimum value.
The LOC can be determined at any stated conditions of
pressure and temperature provided it allows for all possible
concentrations of fuel and diluent. For example, Figure 2
shows ethylene’s LOC at 608C and 16 psia (1.10 bara). The
test conditions were intended to simulate a typical Gulf Coast
vent header [8]. Ethylene’s LOC was found to be 8 mol %
oxygen rather than the classical LOC value of 10 mol %. This
was despite using oxygen analysis plus a robust 100% pres-
sure rise criterion to denote ignition. Furthermore, testing
under identical test conditions had given methane’s LOC as
12% oxygen, the same as the classical value. The apparent
decrease of ethylene’s LOC couldn’t be accounted for by the
elevated temperature of 608C and marginally elevated pres-
sure. Publication of the findings led to the further investiga-
tion of ethylene’s LOC as discussed earlier.
LOC values are commonly measured and tabulated for
atmospheric pressure conditions although a specified elevated
Figure 2. Ethylene triangular flammability diagram. [Color temperature is required for less volatile fuels. As shown in
figure can be viewed in the online issue, which is available Appendix A, several of the LOC values in NFPA 69 were meas-
at wileyonlinelibrary.com.] ured at 608C. This is the high end of the range considered by
UN/DOT as normal transportation temperatures. It’s also a con-
venient “standard” LOC test temperature for fuels such as low
point represents the oxygen concentration needed to com- molecular weight polar solvents that could otherwise create
pletely burn the LFL concentration of fuel. Since Figure 1 is errors owing to condensation during a test. Another “standard”
plotted as a Cartesian fuel-oxygen flammability diagram, the elevated temperature, which is likewise based on transporta-
“nose” region is underemphasized compared with either tri- tion regulations and the NFPA 704 “Flammability Hazard Class”
angular flammability diagrams or those that depict the vol- demarcations, is 938C (2008F). The use of “standard” elevated
ume of added nitrogen needed to inert various temperatures is encouraged to facilitate tabulation and compar-
concentrations of fuels in air. ison of LOC values.
By definition, the LOC is independent of the concentra- For application to partial oxidation reactors, the term
tions of fuel and diluent so there are actually two possible “MOC” is used when operating with a specified minimum
LOC locations, the “lower” near the LFL concentration of fuel fuel concentration. It differs from “LOC” insofar as the fuel
and the “upper” at the UFL in oxygen. For example, carbon concentration is constrained and only diluent concentration
is an independent variable. With a specified minimum fuel
monoxide’s UFL in oxygen is about 95%; since the diluent
concentration, a process can be designed to operate in the
concentration is zero for fuels in oxygen, the “upper” LOC is
nonflammable region above the UFL curve as shown in Fig-
about 5% oxygen regardless of the type of diluent. The
ure 2. Partial oxidation processes are typically controlled by
“upper” LOC is increasingly likely to be the lower of the two
oxygen concentration and the operating envelope is con-
values when the diluent is more effective than nitrogen trolled above the pertinent MOC range using appropriate
(such as H2O or CO2) and the fuel has a high UFL. The safety factors. When the fuel concentration is constrained at
state-of-the-art LOC tests described earlier [4] did not include a relatively high value as required for commercial processes,
investigation of the “upper” LOCs for hydrogen or CO in the MOC greatly exceeds the LOC and the two terms cannot
oxygen so the results only strictly apply to the “lower” LOC be used interchangeably. The MOC can be a function of
such as fuel mixtures with air. multiple time-dependent process conditions, in which case
The existence of an “upper” LOC is implied by Figure 41 calculation of this variable may need to be programmed into
of Kuchta [7], which shows a flammability diagram for mix- the digital control system using a complex polynomial or
tures of hydrogen and air with either N2 or CO2 diluent. As other equation derived from extensive test work.
expected, the LOC was shown to occur at a fuel concentra- Paragraph 7.7.2.6 of NFPA 69 presently exempts “partial oxi-
tion near the LFL although only a marginally greater oxygen dation processes” from operating under the low oxygen
concentration was required at the UFL in air. Although not requirements of Paragraph 7.7.2.5, which would otherwise
discussed by Kuchta, various mixtures of fuels with enriched make such processes unviable. However, the processes could
air or oxygen may exhibit an “upper” LOC that is less than arguably be exempted entirely from Subsection 7.7.2 “Systems
the “lower” LOC, especially when the diluent is more effec- Operated Below The Limiting Oxidant Concentration (LOC),”
tive than nitrogen. UFL data for fuels in oxygen are scarce since during steady operation they are better described by Sub-
but a few additional examples together with their approxi- section 7.7.3 “Systems Operated Above The Upper Flammable
mate “upper” LOC values are: methanol (7%), acetaldehyde Limit (UFL).” The LOC is only relevant to partial oxidation
(7%), and trichloroethylene (9%). The UFL data were taken processes during procedures such as start-up after opening
from Table 6 of Kuchta [7]. Using current ASTM closed vessel equipment to the air; during steady operation, the LOC
test methodology, it’s expected that the UFL values would be becomes irrelevant and the process is controlled below the
increased leading to smaller “upper LOC” values. In the case prevailing MOC.
of “electronic” gases such as dichlorosilane, both the “upper”
and “lower” LOC values are so small that they would be dif- Preferential Diffusion at the LOC
ficult to differentiate experimentally. Discussion of flamma- The “lower” LOC occurs at the optimum fuel/oxygen ratio
bility diagrams is resumed later. for upwards flame propagation. The optimum mixture may

Process Safety Progress (Vol.35, No.1) Published on behalf of the AIChE DOI 10.1002/prs March 2016 109
be greater or less than stoichiometric depending on the rela- some minimum temperature rise recorded by sensitive ther-
tive diffusivities of fuel and oxygen. As summarized by Lewis mocouples. The following year these and other ideas were
and von Elbe [5]: “the more diffusive constituent (either fuel published by Brandes and Ural [11].
or oxygen) is always stoichiometrically deficient.” The stand- For laboratory testing, it’s important to match the over-
ard binary diffusion coefficient of oxygen in nitrogen at 1 pressure criterion adopted with the size of the test vessel,
atm and 298 K is 0.22 cm2/s. Normal butane with a diffusion the type of ignition source and other factors. After the igniter
coefficient of 0.10 cm2/s is much less diffusive than oxygen is fired, there’s always some combustion in excess of the
and its flammability diagram shows an upturned “nose” at heat released by the igniter, even if the test mixture is sub-
the LOC. In this case, the optimum fuel concentration occurs stantially outside the flammable envelope. An overdriven
at 140% of stoichiometric. Oxygen diffuses into the flame flame may propagate a significant distance before self-
front and maintains optimum (stoichiometric) conditions extinguishing and this results in an overpressure that in a
locally despite the overall mixture being fuel rich. Con- laboratory test vessel can amount to several percent of the
versely, hydrogen with a diffusion coefficient 0.78 cm2/s is initial pressure. Even if no hot flame is apparent, reaction
far more diffusive than oxygen and the optimum mixture may nevertheless occur in the rising plume of hot gas mix-
occurs at only 70% of stoichiometric. ture above the igniter. Whenever this is significant, the pres-
When the fuel and oxygen diffuse at similar rates, for sure rise corresponding to mixtures just outside the
example, carbon monoxide, ethylene, or methanol, the effect flammable envelope should be set as a DP “baseline” to
of “preferential diffusion” is marginal although the LOC which the standard 7% pressure rise is added. The sum
“nose” may still be observed for upwardly propagating might result in a practical pressure rise criterion of about
flames. Complex interactions involving preferential diffusion 12% or more.
plus gravitational effects are described by Lewis and von The 120-L state-of-the-art LOC tests were run at normal
Elbe [5] with respect to the measurement of upward flame atmospheric conditions using test gases having large ignition
propagation limits. temperatures (CO, H2, CH4, C2H4, and C3H8); additionally,
the large size of the 120-L vessel minimized background
Possible Effects of Turbulence pressure effects caused by overdriven flames. These particu-
Owing to preferential diffusion effects, the oxygen con- lar test gases are less prone to overdriving than fuels with
centration at the flame front is governed by local mixing and small ignition temperatures such as oxygenated solvents.
differs from the overall gas composition or “measured LOC.” Large and highly variable “baseline” pressures are exhibited
Consequently, one might expect low levels of forced turbu- by various alcohols, aldehydes, ketones, and ethers, particu-
lence to alter the measured LOC by affecting local mixing at larly if tests are made in relatively small vessels at elevated
the flame front. If the lowest LOC values are obtained under temperatures and pressures. In such cases and especially
quiescent conditions, forced turbulence should generally where cool flame phenomena occur, it might be necessary to
increase the measured LOC. If so, one should minimize set the pressure rise criterion between 30 and 100%.
flow-induced turbulence during standard LOC tests using a Although not prominent, “Note 1” of ASTM E2079 allows
large, unstirred apparatus with central ignition rather than a users to adopt a larger pressure rise criterion where neces-
closed, vertical tube ignited at the lower end. A more inter- sary and references Britton [12] who made a number of rec-
esting outcome is if forced mixing typically decreases the ommendations after working with oxygenated fuels
LOC and brings the measured value into closer alignment including aldehydes.
with the value estimated using the approximation that “the
LOC is that concentration of oxygen that will completely Errors in LOC and MOC Measurement
combust the LFL concentration of fuel.” This condition is Since the LOC involves a ternary mixture, it’s inherently
given by the equation LOC5LFL3 S and is represented by more prone to measurement error than either the LFL or the
the point P shown in Figures 1a and 1b. UFL, especially when mixtures are made up by partial pres-
sures without oxygen analysis to confirm the calculated com-
The 7% Solution position. The following list should help improve the poor
Over the years, various pressure rise criteria have been reproducibility sometimes found for gas flammability data
proposed for identifying an “ignition” when testing for flam- obtained using the same standard methods:
mable limits in closed vessels [9]. A 7% pressure rise criterion
was originally proposed because 1 psig approximates to the Pressure Rise Criterion Is Too Small
overpressure threshold for breaking window glass. Also, a ASTM E2079-07 allows test vessels as small as 4 L with an
7% overpressure exceeds the background pressure rise aspect ratio  2 to be used for LOC testing with an ignition
caused by firing igniters in typical laboratory test vessels con- pressure rise criterion of 7%. As discussed earlier, ASTM
taining common gas mixtures below their LFLs. Although a E2079 allows a larger pressure rise criterion to be used
fixed value fails to address the wide range of limit flame tem- where appropriate.
peratures and molar expansions exhibited by disparate fuels,
the 7% pressure rise criterion seemed to be a conservative Ignition Source Is Weak
yet reasonable solution and was adopted by ASTM. In 2005, Energy is usually dissipated inside closed test apparatus
members of the ASTM E27 “Hazard Potential of Chemicals” using a fuse wire, since, unlike spark igniters, these can be
committee published an article discussing the evolution of used at high pressure and aren’t easily shorted out. The
flammability test standards including the new LOC standard exploding fuse wire should dissipate a high power density.
ASTM E2079 [10]. Although energy dissipation from fuse wires isn’t as fast or
In April 2007, the ASTM subcommittee responsible for gas as localized as from a high voltage spark these limitations
flammability standards held a teleconference to discuss vari- are offset by using energies of tens of Joules. The stored
ous problems being experienced when applying the 7% energy needs to be dissipated within the rise time of the
pressure rise criterion. Several proposals were made includ- flame kernel, which can be as short as a few milliseconds.
ing the use of a pressure rise criterion that decreased with Much of the stored energy can be lost in the circuit unless
increasing vessel size, an additional criterion of “cusp” the power cables and connectors are heavy-gauge, low-
behavior in a plot of concentration versus pressure rise and resistance types designed for high currents. Higher

110 March 2016 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.35, No.1)
resistances and capacitances increase the RC time constant of connected lines and crevices such as strain gauge casings
the circuit and can lead to extended discharge times. The and even if the vessel is stirred, which is often not done at
combination of long discharge times, large circuit losses, and all, trapped liquid is mostly unaffected. The result is usually
a reduced oxygen atmosphere may result in the wire not an underestimate of the concentration of oxidant since the
exploding but instead functioning as a glowing wire. The target pressure is based on the initial measurement of fuel
time constant is decreased using higher voltages, shorter partial pressure. In an extreme case, it might be impossible
cables, and lower storage capacitances. The 5.3-L apparatus to make steady pressure measurements until all the adventi-
described in [8,12] was equipped with switched parallel fuse tious cold traps are filled with liquid. To minimize condensa-
wires that were each designed to dissipate 30 Joules of tion problems, Britton [12] used a twin vessel method with
energy within 5 ms, as confirmed by high speed digital V-I both vessels placed inside a convection oven, including all
measurements across the igniter itself. Prior to upgrades in connected lines and gauges. Only the oxygen analyzer was
the 1980s most of the stored energy had been lost heating outside the oven and the sampling line was extensively heat
the power cables and much of what remained might have traced. In general, the test temperature must exceed the
been dissipated wastefully, after flames had already become value calculated to produce the required vapor pressure of
self-sustaining. fuel. Standard elevated temperatures considered by ASTM
2079 are 60 and 938C.
Compressed Air Isn’t “Air”
It’s important to define the LOC and other flammable lim- No Oxygen Analyzer
its on a dry gas basis since, depending on temperature and The previous two sources of error can be mitigated using
relative humidity, atmospheric air can contain several mole an oxygen analyzer. To obtain a reliable reading from the
percent of water vapor. Since water vapor would otherwise test vessel using a paramagnetic analyzer or other flow
act as an inert component, testing on a dry gas basis pro- instrument, the test mixture must be made up at an elevated
duces conservative results. Although argon (0.93 mol %) is pressure. For potentially condensable liquids, the fuel vapor
usually neglected, it’s effectively grouped with nitrogen pressure needs to be considered in relation to the test tem-
when oxygen is specifically analyzed. Compressed air is usu- perature. Oxygen analysis isn’t required by ASTM E2079
ally not “standard” air; for example, CGA Grade D com- although a single confirmatory oxygen analysis is required
pressed air cylinders (OSHA breathing air) can contain by EN 14756 [13] and this value must agree to within
between 19.5–23.5 mol % oxygen plus up to 0.1 mol % car- 60.2 mol % of the calculated concentration. As a practical
bon dioxide. Although CGA Grade J (specialty/analytical matter, if paramagnetic oxygen analysis is used to confirm
grade) has much tighter restrictions on trace gases, it con- the calculated oxygen concentration in the test mixture, the
tains the same range of oxygen concentrations as Grade D. initial pressure must be adequate to generate the required
Dry “synthetic air” made up for test purposes should contain flow rate and sampling time of the analyzer. For 5-L test ves-
the “standard” concentration of oxygen in air sels, an initial pressure of 25 psia (1.7 bara) might be neces-
(20.94 6 0.1 mol %) as required under Section 6.3 of ASTM sary and the temperature must be adequate to maintain the
E681-09 (2015). This requirement is currently absent from fuel as a vapor during this sampling period. Since paramag-
ASTM E 2079. Moreover, this requirement is missing from netic analyzers respond to oxygen partial pressure, the sen-
two of the five ASTM standards that address the flammability sor unit is maintained at a fixed pressure and temperature. If
of dust clouds. the sample flow rate and temperature aren’t properly
selected, the analyzer will give erroneous readings.
Wrong Strain Gauge
When making up mixtures using the method of partial Cool Flames Occur
pressures a sensitive strain gauge must be used rather than Britton [12] showed that at elevated temperatures and
the high range strain gauge used to measure the deflagration pressures, it’s impractical to use a 7% pressure rise as an
pressure rise. Calibrations can drift so both zero and span ignition criterion for gases subject to cool flame phenomena,
should be routinely checked. Also, if the gauges are exposed whether or not this is added to a small “baseline” DP value.
to elevated temperature they must be temperature On fuel-oxygen flammability diagrams the phenomena may
compensated. form a “second lobe” on the fuel-rich side of the LOC “nose.”
Calculations showed gas compositions in this “second lobe”
Mixture Isn’t Mixed had adiabatic flame temperatures less than 1,000 K which are
ASTM E2079 doesn’t require the test vessel to be inadequate to support hot flame propagation. Recommenda-
equipped with a mixer or gas analyzer although it requires tions for avoiding these problems were later incorporated
“uniform mixing of the components before attempting into ASTM E2079-07. The significance of cool flames is that
ignition.” However, the 120-L test vessel used by Zlochower the reaction might transition to a two-stage ignition or an
and Green [4] was equipped with an internal fan and sam- autoignition. However, although autoignition is not pre-
ples were taken to confirm that thorough mixing was accom- vented by operating below either the LOC or MOC, a
plished. Smaller vessels such as the 20-L apparatus are flame by definition cannot propagate into properly inerted
difficult to retrofit with a mixing device but lack of a mixer equipment connected to the zone where such phenomena
may cause errors, especially when testing liquids or vapors occur [12].
subject to condensation during testing.
Irreproducible Positive Results
Fuel Condenses ASTM E2079 requires the LOC to be calculated as the
After adding the test liquid and allowing it to evaporate, average of the lowest oxygen concentration at which ignition
oxidant and diluent gases are added from cylinders and the occurs L and the highest oxygen at which ignition doesn’t
cold gas jets often re-condense some of the vapor. Addition occur H. In case the lowest observed L is wrong owing to
of diluent is sometimes used to assist mixing, in which case one of the many possible errors, ASTM E2079 provides that
it’s done last. Condensation decreases the fuel in the vapor at the LOC the difference between H and L should not
phase and to reach the target pressure more gas is added exceed 0.5 mol % oxygen. In some test standards, a single
resulting in too much oxidant or diluent depending on the positive result is allowed to outweigh any number of nega-
order of addition. Condensed liquid can become trapped in tive results with no requirement for even a single

Process Safety Progress (Vol.35, No.1) Published on behalf of the AIChE DOI 10.1002/prs March 2016 111
confirmation! Using ASTM E2079, if H 2 L exceeds 0.5 mol should, therefore, predict the LOC conservatively in all cases,
% oxygen and the gap cannot be narrowed with more test- provided the LFL value is reliable.
ing, the lower value can be discarded.
LOC5LFL3S (1)
Complex Liquid Fuel Mixtures
Commercial liquid fuels are better characterized by distil- According to Eq. 1, the location of the LOC is at point P
lation ranges than by flash points; this is due to the wide on Figures 1a and 1b. Where identical test methods are used
boiling point range of the components (which may run into to measure LFL and LOC, Eq. 1 typically underestimates the
the hundreds). In such cases, it’s neither practical nor appro- LOC by up to about 2 mol % oxygen. This conservative
priate to test the entire mixture. For applications such as result is usually satisfactory because a safety factor of at least
inerted naphtha, kerosene, or diesel storage tanks, a suitable 2 mol % oxygen is required by NFPA 69 and very precise
LOC value is normally assumed rather than measured. Even LOC values are typically not needed. In some cases, Eq. 1
if commercial fuel mixtures contain aromatic components, overestimates the LOC, but usually this is due to the assumed
inerting to below 5–6 mol % oxygen usually meets NFPA 69 LFL value being too high. There are a few cases (such as CO
requirements. If a representative LOC value is necessary, a as shown in Figure 1b) where the LOC is apparently overes-
timated by about 1 mol % oxygen but this is less than the
specialized test needs to be conducted so that minor compo-
minimum applied safety factor.
nents in the liquid phase are not under-represented in the
Britton [8] cited Bodurtha [15] as the source of Eq. 1 and
vapor phase once the system has achieved vapor-liquid equi- commented that a likely source of error when applying the
librium. The problem of providing an adequate liquid reser- equation is that any error in LFL becomes multiplied by S,
voir is discussed by Britton and Smith [14] in relation to which for aromatics and other fuels with relatively high
storage tanks at high elevation, where the lower prevailing molecular weights can have a value up to 10 or more. For
pressure significantly decreases the “flash point.” Numerous example, styrene is easily condensable and has a small LFL
unexpected land-based diesel and kerosene tank fires have (0.9 mol %) which could easily be overestimated during a
occurred at high altitude. Perhaps the most critical applica- test. The LFL commonly given in the literature is 1.1 mol %.
tion is the LOC of aviation fuels which must address the Application of Eq. 1 multiplies the error by 10 and the LOC
worst case with respect to both temperature and pressure. (9 mol %) is estimated as 11 mol %. This type of error is
Special large scale tests are described in the literature. illustrated by Table 2 of Hansen and Crowl [16].
When the fuel concentration at the LOC exceeds stoichio-
Fuel Reacts Rapidly with Oxidant metric, the LOC is effectively a rich limit, with excess fuel
For some process applications involving reactive gas mix- acting as a second diluent. Were the excess fuel partially oxi-
tures, the LOC or MOC must be measured in a flow system. dized to CO, the assignment of S would become compli-
Aromatics with chlorine oxidant are particularly difficult. cated. However, one of the few fuels whose measured LOC
Testing techniques are highly specialized and beyond the is less than the product LFL3 S is CO itself, which cannot be
scope of this article. partially oxidized in the flame. Since the excess CO contrib-
utes no heat, its involvement must be to enhance transport
“Experimental Error” Versus “Definition of Ignition properties despite the similarity in the diffusion rates of CO
Error” and oxygen. The oxidation rate of CO is slow unless a
There are two sources of discrepancy between classical source of hydrogen is present—in which case the fuel system
LOC values and current LOC values. Of these, the definition is referred to as “moist” CO. Hydrogen acts as a catalyst and
of ignition error (propagating flame in tube vs. 7% pressure the rate of oxidation of CO is controlled not via reaction
rise in closed vessel) might account for up to about 10% of with oxygen but with OH radicals. Figure 1b, adapted from
the absolute LOC (for example 1 mol % oxygen when the Figure 13 of Coward and Jones [6], shows that the LOC of
LOC is 10 mol % and 0.5 mol % oxygen when the LOC is moist CO occurs at about 17% fuel or 4–5 mol% above the
5%) and is generally negative, so that “new” LOC values are LFL concentration. The original Coward and Jones flammabil-
less than classical values. Assuming the new LOC values are ity diagram shows the “nose” region far more clearly than
Figure 1b since it plots CO concentration versus “nitrogen
correct and that experimental errors in classical data account
added to the original atmosphere” rather than oxygen
are normally random (such as composition errors) rather than
concentration.
biased (such as weak ignition errors), the discrepancy caused
by experimental errors may or may not be additive to the def-
Effects of Elevated Temperature and Pressure
inition error. For a particular igniter, the repeatability of ASTM
Equation 1 can be applied to elevated temperatures by
E2079 should be within 0.5 mol % oxygen so experimental
adjusting the LFL value. Procedures for correcting the LFL
errors associated with classical LOC determinations should not using a linear interpolation technique are discussed by Brit-
be better than 60.5 mol % oxygen. With respect to the recent ton and Frurip [17]. The technique requires the lower limit
large-scale ASTM tests, the average observed discrepancy was flame temperature to be known or estimated.
about 7.5% of absolute LOC but increased to about 15% for As discussed by Britton and Frurip [17], although the
ethylene. The latter case could correspond to a condition effects of elevated pressure on LFL are relatively small, Eq. 1
where all the errors were large and additive. predicts that the effects of pressure on LOC will be S times
Additional testing in a 5-cm vertical tube would determine as large. Consequently, while the effect on LFL might be less
whether or not the discrepancy found for ethylene was due than experimental error (0.1–0.2 mol % fuel) the effect on
to flame quenching. If ethylene cannot propagate, a flame LOC for relatively high molecular weight fuels might be of
below 10% oxygen using an adequate ignition source the the order 1 mol % and, therefore, significant.
quenching thesis will have been validated, and vice versa. Figure 41 of Zabetakis [18] shows that for natural gas, eth-
ane, and propane, the LOC decreases with increasing pres-
LOC Estimation sure, although the gradients are relatively small. Britton and
If complete combustion occurs and if the LFL does not Frurip [17] observed that one practical effect of elevated
decrease as diluent is added, the LOC cannot be smaller than pressure would have been to offset heat losses in the small
the product of LFL and the stoichiometric ratio S. Equation 1 diameter test vessels used and this should account for at least

112 March 2016 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.35, No.1)
part of the observed LOC decrease with increased pressure. tends to smooth out individual LFL errors in the literature
It was also observed that as pressures were increased up to plus the effects these would have on the estimated LOC.
about 20 bar the measured LOC values tended toward those
predicted by Eq. 1. CONCLUSIONS
ASTM’s CHETAH program is discussed in Appendix B. Flammable limits are used not only in safety analysis but
The program allows LFL values to be estimated for many also in combustion science. In both cases, it’s important that
organic fuels, either at a standard temperature of 298 K or at limits measured in the laboratory agree closely with the lim-
a user defined elevated temperature. The LOC is estimated its observed in large volumes of gas. Although wider flam-
from this LFL using Eq. 1. Pressure is not addressed and mable limits might be claimed to be “more conservative” it’s
without additional information on the topic it’s assumed that, not the job of a test laboratory to include safety factors in
at least within certain limits, the effect of increased pressure the test method. Such safety factors should be included dur-
is marginal. Elevated pressure may actually increase the ing application of data, not during measurement. With
accuracy of Eq. 1 since this is typically conservative for respect to closed vessel tests, having lost sight of the flame
organic fuels at atmospheric pressure. it’s important not to lose sight of the objective, namely to
select test criteria that correctly predict the flammable limits
Estimated LOC Values for Mixtures pertaining to large volumes of gas mixture [9].
Proposed TIA changes to NFPA 69 involve (1) revisions to Since these sentiments were published in 2002, European
Annex Table C.1(a) and (2) a new section explaining how to test methods have continued to generate flammable limit data
estimate the LOC of individual fuels or fuel mixtures. The outside the envelope of compositions within which flames
TIA proposals are detailed in Appendix A of this article. can propagate through large volumes of test mixture. Mean-
Methods for estimating the LOC of fuel mixtures differ while, although the 7% pressure rise “ignition criterion” used
depending on whether or not LOC values are available. Even in ASTM E2079 has been shown to be valid during 120-L
if only LFL values are available the LOC of mixtures can be “reference quality” tests, the scope of these tests was limited
estimated. Methods exist for estimating LFL and hence LOC to only five gases under normal atmospheric conditions. Also,
at elevated temperature. it’s been found that the 7% pressure rise criterion must often
Prior to the general acceptance of these LOC mixture esti- be increased when using smaller test vessels, especially at ele-
mation methods, many users have designed explosion protec- vated temperatures and pressures. The need for a larger pres-
tion based on controlling below the smallest LOC of the sure rise criterion under certain conditions is recognized by
component materials, clearly wasting a good deal of inerting ASTM E2079. Not only is there a provision for a “baseline”
gas. Such waste is especially severe for large throughput waste pressure rise to which the standard 7% value is added, but
combustors such as flares, where headers are run under inert the pressure rise criterion can be substantially increased based
conditions rather than fuel-rich conditions. A large underesti- on “exploratory tests in the vicinity of limit mixtures.” How-
mation of mixture LOC will in such cases lead to the use of ever, since the procedure for selecting nonstandard pressure
excessive amounts of inert gas which could in turn lead to rise criteria is vague, test laboratories may prefer to err on the
additional natural gas input to the flare. The new approach side of a conservative result. Some specific conclusions are:
allows LOC estimates for complex mixtures and provides a sig- 1. Additional tests should be conducted in NIOSH’s 120-L
nificant improvement over the assumption that a gas mixture’s vessel or a similar large test vessel capable of producing
LOC is represented by the smallest LOC of any component. “reference quality” LFL and LOC data. Test gases should
include a variety of “CHON” fuels including examples of
Hybrid Approach fuels with low ignition temperatures. Tests should be
Although many LOC values are tabulated in NFPA 69, the repeated under stirred conditions to address the effect of
list is far from comprehensive. Appendix A describes the mild turbulence, which might increase or (more impor-
general methodology for estimating the LOCs of mixtures tantly) decrease the measured LOC value. The test pro-
starting either with LOC values or LFL values. However, in gram should preferably include elevated initial
some cases a user may have “reliable” LOC values for some temperatures or pressures. The LFL and LOC data would
components but only obsolete or unknown LOC values for be extremely helpful for evaluating LOC estimation tech-
other components of a mixture. In such cases, a “hybrid” niques such as those discussed in this article.
approach may be used that combines the “reliable” LOC val- 2. Additional testing should also be carried out using smaller
ues with LOC values estimated using Eq. 1, which are gener- vessels that better represent those used by testing labora-
ally conservative. The approach is very similar to the one tories; typical volumes are between 5 and 20 L. Such test-
described in Appendix A and involves a temporary conver- ing would help standardize the pressure rise criterion
sion of all the LOC values to “pseudo-LFL” values using Eq. with respect to test vessel volume rather than type of
1. These are then combined using Le Chatelier’s rule and igniter and type of fuel. The latter variables can be sepa-
reconverted to a mixture LOC. rately addressed using the “adjustable baseline” proce-
dure currently described in ASTM E2079.
Elevated Temperatures 3. The simple LOC estimation method based on Eq. 1, that is,
If the LOC is required at an elevated temperature, tabu- LOC5LFL3 S, is generally conservative when applied to
lated LOCs can seldom be used directly. If the fuel compo- individual fuel components. In the few cases where the
nents all consist of carbon and hydrogen, with or without LOC is overestimated, and where this is not due to an erro-
nitrogen and chlorine, ASTM’s CHETAH program can be neously large LFL value, the error appears to be less than
especially helpful. CHETAH currently only allows calcula- the minimum application safety factor required by NFPA 69.
tions for fuels that burn to carbon dioxide and water, with or Where the estimation method is applied to fuel mixtures,
without nitrogen and hydrochloric acid in the products. As most component LOCs will be conservatively estimated and
discussed in Appendix B, CHETAH requires no user input these should normally offset the occasional nonconserva-
apart from the fuel and temperature of interest to provide tive estimate for individual mixture components.
estimates of the LFLs of single fuels and mixed fuels at the 4. Since Eq. 1 multiples any LFL error by S, which may be
user-defined temperature. Equation 1 is used to estimate the as high as 10 or more, it’s important to select realistic LFL
corresponding LOCs. The method employed by CHETAH data. The use of erroneously high LFL data will give LOC

Process Safety Progress (Vol.35, No.1) Published on behalf of the AIChE DOI 10.1002/prs March 2016 113
predictions that are much too high and also nonconserva- 3. ASTM International, ASTM E2079, Standard Test Methods
tive; examples include aromatics and other readily con- for Limiting Oxygen (Oxidant) Concentration in Gases
densable fuel species with large S values. Conversely, and Vapors, ASTM International.
overly conservative LOC values may result in unnecessary 4. I.A. Zlochower and G.M. Green, The limiting oxygen
expense, particularly when the application involves large concentration and flammability limits of gases and gas
volumes of gas such as a vent header going to a flare or mixtures, J Loss Prev Process Ind 22 (2009), 499–505.
other destruction device. 5. B. Lewis and G. von Elbe, Combustion, Flames and
5. Apart from ethylene, whose “classical” LOC value is consid- Explosions of Gases, Elsevier, 2012.
ered to have been the result of flawed measurement rather 6. H.F. Coward and G.W. Jones, Limits of Flammability of
than a systemic “quenching” problem with the 5-cm vertical Gases and Vapors, Technical Report 503, U.S. Bureau of
tube apparatus, much of the difference between classical Mines, 1952.
LOC values and values measured using ASTM E2079 can be 7. J.M. Kuchta, Investigation of Fire and Explosion Acci-
attributed to different definitions of “ignition.” Owing to its dents in the Chemical, Mining, and Fuel-related Indus-
more conservative definition of “ignition” ASTM E2079 has tries, Technical Report 680, U.S. Bureau of Mines.
a negative bias relative to the 5-cm vertical tube apparatus 8. L.G. Britton, Operating atmospheric vent collection head-
so values obtained using ASTM E2079 should normally be ers using methane gas enrichment, Process Saf Prog 15
less than those found by classical methods. (1996), 194–212.
6. LOC values presented in Table C.1(a) of Annex C of the 9. L.G. Britton, Two hundred years of flammable limits, Pro-
last two editions of NFPA 69 contained classical LOC val- cess Saf Prog 21 (2002), 1–11.
ues that had each been adjusted downwards by 2 mol % 10. L.G. Britton, K.L. Cashdollar, W. Fenlon, D. Frurip, J.
oxygen, regardless of the initial magnitudes. The adjust- Going, B.K. Harrison, J. Niemeier, and E.A. Ural, The role
ment was driven primarily by the reported behavior of of ASTM E27 methods in hazard assessment part II: Flam-
ethylene. Proposed revisions to Table C.1(a) are dis- mability and ignitability, Process Saf Prog 24 (2005), 12–
cussed in Appendix A of this article. The proposals first 28.
replace classical LOC data with recent 120-L vessel LOC 11. E. Brandes and E.A. Ural, Towards a global standard for
data obtained using ASTM E2079. They subsequently flammability determination, Proceedings of the 42nd
reduce the overcorrection of the previous two editions of Annual Loss Prevention Symposium–Global Safety Con-
NFPA 69 by limiting the downward LOC adjustment to
gress, Paper 2E, April 6, 2008.
the smaller of either a 1.5 mol % reduction or a 15%
12. L.G. Britton, Flammability hazards of lower aliphatic alde-
reduction of the classical LOC value. The same adjust-
hydes at elevated pressure and temperature, Process Saf
ments are applied for both nitrogen and CO2 diluent gas,
Prog 17 (1998), 138–148.
although there’s presently no 120-L vessel data for an
13. European Committee for Standardization, EN 14756,
adjustment of the classical values with CO2 diluent.
Determination of the Limiting Oxygen Concentration
7. Appendix A discusses simple estimation techniques for the
LOC of mixtures. The estimates can begin with LOC values, (LOC) for Flammable Gases and Vapours, European
LFL values, or a mixed set containing only some LOC values. Committee for Standardization, Brussels, 2006.
The estimates can be made for elevated temperature condi- 14. L.G. Britton and J.A. Smith, Static hazards of the VAST,
tions using corrected LFL values. The procedures have been J Loss Prev Process Ind 25 (2012), 309–328.
proposed as revisions to NFPA 69. 15. F.T. Bodurtha, Industrial Explosion Prevention and Pro-
tection, McGraw-Hill, New York, 1980.
16. T.J. Hansen and D.A. Crowl, Estimation of the flammabil-
ity zone boundaries for flammable gases, Process Saf
LITERATURE CITED Prog 29 (2010), 209–215.
1. P. Yanisko, S. Zheng, J. Dumoit, B. Carlson, et al., Nitro- 17. L.G. Britton and D.J. Frurip, Further uses of the heat of
gen: A security blanket for the chemical industry, Chem oxidation in chemical hazard assessment, Process Saf
Eng Prog 107 (2011), 50–55. Prog 22 (2003), 1–19.
2. National Fire Protection Association, NFPA 69, Standard 18. M.G. Zabetakis, Flammability Characteristics of Combusti-
on Explosion Prevention Systems, National Fire Protec- ble Gases and Vapors, Technical Report 627, U.S. Bureau
tion Association, Quincy, MA, 2014. of Mines, 1965.

114 March 2016 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.35, No.1)

Das könnte Ihnen auch gefallen