You are on page 1of 15

New Model for Predicting Thermal Radiation from

Flares and High Pressure Jet Fires for Hydrogen


and Syngas
Derek Miller
Chief Engineer’s Office, Air Products and Chemicals, Inc, Allentown, PA; Millerd3@airproducts.com (for correspondence)
Published online 00 Month 2016 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/prs.11867

Current flare and jet fire models used in the process [3] with the new AP Flame model. The Chamberlain model is
industry are primarily based on hydrocarbon gases and really only for light hydrocarbons, but is often used more broad-
have been found not to be particularly effective for low lumi- ly, API 521 is for light hydrocarbons and hydrogen. Figure 1
nosity gases such as hydrogen and syngas mixtures. This shows results for a typical hydrogen and syngas flare, across a
article presents a new model for such gases, for low to high range of conditions up to choked flow. The example flare is actu-
pressure releases in both vertical and horizontal orientations. ally located at the Air Products Garyville hydrogen plant, where
The model, which predicts flame geometry and thermal radi- tests were performed at turndown rates for both hydrogen and
ation, allowing for jet momentum, buoyancy, wind and syngas (73% H2, 7% CH4, 4% CO, and 16% CO2) [4]. Terminology
flame radiant fraction, is based on extension of established is described in third section of this paper.
models, with the addition of several new correlations. Valida-
tion against existing published data and new previously
unpublished test data collected at both pilot and full scale is API 521
presented. This article provides the complete set of equations API 521 matches hydrogen flare test data points quite
for the model. V C 2016 American Institute of Chemical Engineers
well, but overestimates radiant fraction and flame length at
Process Saf Prog 000: 000–000, 2016 high rates, as choke flow is approached. API 521 model is
Keywords: hydrogen; syngas; jet fire; flare; model only valid for subsonic flares. API 521 overpredicts flame
length and tilt for syngas.
OVERVIEW
Hydrogen and syngas (mixtures containing hydrogen, car- Chamberlain
bon monoxide, and other components) are routinely used It should be noted that this model was intended for use
across a wide range of industrial applications. One important with natural gas and was never intended for use with hydrogen
hazard to manage is jet fires, both planned (e.g., flares) and or syngas. Nonetheless, it matches hydrogen test data points
unplanned (e.g., process leaks). There is currently no pub- quite well, but overestimates radiant fraction at lower rates. Air
lished model that adequately predicts flame geometry and Products has observed flames getting longer at high tilt angles,
thermal radiation for these gases across a wide range of gas (which has caused operational issues) but this is not picked up
compositions and release conditions (velocity, temperature, by Chamberlain. To compound this issue, hydrogen flames are
and orientation), taking adequate account of ambient condi- more greatly impacted by wind than natural gas flames, which
tions, such as wind speed—Air Products previously presented is why Chamberlain underpredicts tilt for hydrogen. Chamber-
a comparison between existing models and published data at lain overpredicts radiant fraction and tilt for syngas.
GCPS in 2014 [1]. A new model is presented here, built on
extension of published models primarily developed for hydro-
carbon flames, which has been validated against available pub- AP Flame
lished data, as well as new test data specially collected for this AP Flame matches test data points well (expected since
purpose. This report provides the equations required for the data are part of overall dataset). General graph trends sup-
model with the available experimental data and validation ported by results from other tests in dataset.
between data and model. The new model is called AP Flame.
TERMINOLOGY
COMPARISON OF MODELS Figure 2 illustrates key terms used in this article. AP Flame
To introduce the subject, it is worthwhile to briefly compare models vertical, horizontal, and 458 releases in the wind
results of two established models, Chamberlain [2] and API 521 direction.

This article is prepared for presentation at American Institute of Coordinates


Chemical Engineers 2016 Spring Meeting, 12th Global Congress on
Process Safety Houston, Texas, April 10–14, 2016. X is in wind direction. Y is in the vertical direction. Z
is in the sideways direction, perpendicular to the wind
C 2016 American Institute of Chemical Engineers
V direction.

Process Safety Progress (Vol.00, No.00) Month 2016 1


Figure 1. Comparison of model results: Chamberlain, API 521, and AP Flame for 2700 diameter, 53 m high vertical hydrogen
and syngas (73% hydrogen) flare, rural topography. [Color figure can be viewed at wileyonlinelibrary.com]

Flame Dimensions lift off from the flare tip (no combustion at tip). For purposes
Initial momentum section length, bl: Flame initially travels of this model, bl is the total length that is momentum dominat-
straight in release direction due to momentum. For some ed, included any lift off. Wind–buoyancy section length, RL:
gases with relatively low burning velocities, the flame may flame travels in direction determined by buoyancy and wind

2 Month 2016 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.00, No.00)
Figure 2. Flame geometry. [Color figure can be viewed at wileyonlinelibrary.com]

Table 1. Summary of data sources.

Release Flare Release Flame


Data Source Gas Direction Diameter (”) Rate (kg/s) Length (m) Test Location Reference
Fishburne, 1979 H2 Vertical 6–27 2–126 48–110 USA Rocketdyne [6]
test site
Chamberlain, 1987 NG Vertical 6–42 4–63 – UK on & off shore [2]
Shirvill, 2005 H2 Horizontal 0.12–0.24 0.02–1 4–6 UK HSL site [7]
Schefer, 2007 H2 Vertical 0.2–0.3 0.06–0.4 – USA SRI test site [8]
DNV GL, 2008 NG and NG/H2 Horizontal 0.8–2 3–20 20–50 UK Spadeadam [9]
mixtures
Willoughby, 2011 H2 Vertical 0.75–2 0.04–0.4 3–7 UK HSL test site [10]
Air Products, 2014 H2 Vertical 36 1.3 30 Pasadena, TX H2 plant [1]
Air Products, 2014 H2 and syngas Vertical 27 1.7–13 25 Garville, LA H2 plant [4]
Air Products/ H2 Horizontal 0.7–2 1–7.5 17–49 UK Spadeadam Previously
DNV GL, 2009 unpublished
Air Products, 2015 NG, H2, Horizontal 1 0.02–0.07 3–5 USA R&D lab Previously
mixtures unpublished
with N2, CO2

conditions. Vertical flame direction defined by tilt angle, a. heat emitted and the distance between the source and the
Horizontal flames defined by lift, Ly or lift angle, d. Centerline receiver, Ri), with allowance for the air transmissivity, the
flame length, Lf is the overall flame length (bl 1 RL). fraction of thermal radiant passing through a thickness of air
(not absorbed). Heat fluxes from other thermal radiation
Flame Radiation sources, such as solar, should be added to determine total
The fraction of the total heat released in the flame that is heat flux at any point of interest.
in the form of thermal radiation is called the radiant fraction,
F determined empirically from test data. The total heat radiat- SUMMARY OF DATA
ing from the flame in AP Flame is distributed as individual Table 1 summarizes all data sources used in this analysis.
point sources along the flame centerline, weighted to be a Full set of data used in the development of AP Flame is pro-
maximum at the widest point in the flame (about 2/3 along vided at the end of this article.
the flame, the radiant section), with low contributions from
the momentum section and at the tip, where the flame is Previously Published Data
wispy [5]. The heat flux experienced at the receiver is the Fishburne (1979) [6] performed large-scale flare testing of ambi-
sum of the heat received from each point source (function of ent and cold hydrogen vertical releases through a variety of vent

Process Safety Progress (Vol.00, No.00) Published on behalf of the AIChE DOI 10.1002/prs Month 2016 3
Figure 3. Typical flames observed during the various Air Products tests. [Color figure can be viewed at wileyonlinelibrary.
com]

sizes. Flame length and radiant fraction results provided. Chamber- radiant fraction to develop methods to model syngas mixtures.
lain (1987) [2] performed large-scale vertical natural gas (and simi- Results of these tests are presented in this report.
lar) flare testing at off-shore and on-shore facilities and presented
a flare model derived from this data which is used as the basis for Additional Data
the new model presented in this article. Shirvill (2005) [7] per- During the review process for this article, another important
formed tests on high pressure horizontal hydrogen releases, both source of test data was identified. Johnson et al. [13] tested large
ignited and unignited and presented flame length results. Schefer horizontal natural gas jet fires and developed a model, basically
(2007) [8] performed tests with high pressure vertical hydrogen extending the work of Chamberlain to horizontal releases. Includ-
releases and presented flame lengths and radiant fractions, as ed in the Johnson paper are four sets of test data. At the end of
well as a model for high pressure, vertical jet fires in the absence this article, a comparison between Johnson data and AP Flame
of wind. DNV GL (previously Advantica) (2008) [9] performed predictions is provided. Overall, AP Flame performs quite well.
tests on high pressure horizontal natural gas and natural gas/
hydrogen mixtures and presented flame lengths and heat flux TYPICAL FLAMES OBSERVED DURING AIR PRODUCTS TESTS
measurements, from which this author has derived radiant frac- Figure 3 shows a selection of flames observed during Air
tions. Willoughby (2011) [10] performed tests on vertical jet fires Products tests. Figures 3a and 3b were obtained during the
and provided photos of flames, with weather conditions as well Air Products/DNV GL tests at Spadeadam, for hydrogen
as heat flux values at different locations, from which this author releases from 3=400 and 200 horizontal pipe. Figures 3c and 3d
has extracted flame lengths, flame tilt, and radiant fractions. Air were obtained during the Air Products Garyville full scale
Products (2014) [1,4] performed tests on full scale hydrogen plant vertical flare tests for hydrogen and syngas using IR camera.
vertical flares with both hydrogen and syngas (73% H2) and pre- Figures 3e and 3f were obtained during Air Products R&D
sented results on flame length, heat flux, and radiant fraction. tests for 100 458 natural gas and horizontal hydrogen releases.

Air Products Previously Unpublished Test Data AP FLAME BASIS


In 2008, the company commissioned DNV GL (then Advan- AP Flame is based on the Chamberlain model [2], devel-
tica) to perform tests on high pressure horizontal hydrogen jet oped in the 1980s from extensive data collected from natural
fires, collecting information on flame length, heat flux, and gas vertical flares. The new model also includes features
radiant fraction [15]. In 2015, Air Products completed testing at from the model developed by Sandia National Labs [5] for no
its R&D facility on horizontal and 458 jet fires at different eleva- wind high pressure flames, primarily hydrogen. AP Flame
tions, of mixtures of natural gas, hydrogen, nitrogen, and car- includes a number of newly developed correlations to ade-
bon dioxide and obtained data on flame length, heat flux, and quately model mixtures and horizontal/458 release flames.

4 Month 2016 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.00, No.00)
None of the modifications impact the predictions for natural
gas from the original Chamberlain model. All equations from
the Chamberlain model are clearly identified in the text with
“Chamberlain” designation after the equation number. All
other equations represent new developments unless other-
wise referenced. All symbols used are described at the end
of the article. To help follow the series of equations, this is a
simple overview of the calculation steps
1. Determine source condition for the flame, expanded
source velocity uexp, density qexp, and diameter dexp.
2. Determine Richardson number n (LBO) and zero wind
flame length LBO, both used extensively as correlation
variables.
3. Determine flame dimensions bl, RL, Lf, and tilt a (for verti-
cal releases) or lift d (for horizontal releases).
4. Determine flame radiant fraction F based on gas mixture
and release orientation. Figure 4. Adiabatic combustion temperature correlation vs.
5. Determine overall heat flux, q at point of interest, by NASA CEA results. [Color figure can be viewed at wileyonli-
summing heat fluxes received from thermal radiation nelibrary.com]
sources along flame.

JET RELEASE EXPANDED SOURCE CONDITION FOR CHOKED FLOW


RELEASES
It is necessary to determine conditions downstream of the Zero Wind Flame Length, LBO
choke, where pressure is atmospheric, called the expanded (or This parameter is used extensively in the model and Cham-
effective) source condition. The expanded source condition berlain uses a previously developed correlation, equation (6) by
defines the source term used in AP Flame. A number of different Kalghatgi [20] which is solved iteratively to obtain LBO, using the
methods have been proposed to determine the expanded source Richardson number. mwair 5 molecular weight of air, mwpro-
condition, including Birch (1984) [19]; Birch (1987) [17] and ducts 5 molecular weight of stoichiometric combustion products.
Ewan & Moodie [16]. AP Flame uses the Yuceil methodology [11] Tad 5 Adiabatic combustion temperature (constant pressure),
to determine conditions at the expanded source, using conserva- Tair 5 Ambient air temperature. Fs 5 stoichiometric mass frac-
tion of mass, momentum, and energy. Expanded source condi- tion fuel in fuel 1 air (note: simplified equation for Fs provided
tions (uexp, qexp, dexp) are calculated from the release throat in original Chamberlain paper only applicable to pure fuels of
(pipe exit) condition, where isentropic expansion and choked form CnH2 n 1 2 and should not be used for syngas). Determina-
flow, if applicable, are applied, (qexp is determined from Hexp tion of Tad can be done using the following newly developed
and Pexp 5 Patmos). Note that the expanded source Mach correlation, based on Tad values for a range of syngas composi-
number > 1 when throat is choked. For nonchoked releases, tions from pure H2 through to condensate off-gas (15% H2,
conditions at the throat and at the expanded source are the remainder CO2 and water vapor) determined using the online
same, Mach number < 1. The mass release rate, m_ is either speci- Chemical Equilibrium with Applications (CEA) from NASA [12].
fied or calculated based on upstream pressure. The correlation provides a reasonable fit, see Figure 4.
The equations for conservation of momentum, energy,  2=3
and mass are written as (1)–(3), Ds b
50:210:024nðLBO Þ (6)Chamberlain
LBO Fs
Pthroat 2Patmos
uexp 5uthroat 1 (1)
uthroat qthroat Where b is the density correction,
1 2 
Hexp 5Hthroat 2 uexp 2u2throat (2)  0:5
2 mwair Tad
b5 (7)Chamberlain
m_ p mwproducts Tair
5 d2 (3)
qexp uexp 4 exp In which,
FLAME GEOMETRY FUNDAMENTALS
 
Tad 5Tad;fuel 20:7395CD2equiv 10:0366CDequiv 10:9972
Richardson Number, n (LBO)
This dimensionless number is used extensively by Cham- Tad;fuel 5321:08H32equiv 2310:88H22equiv 1144:03H2equiv 12223
berlain and represents the ratio of buoyancy to momentum CDequiv 5xCD 10:84xWA 10:59xN2
flux for the zero wind flame length, LBO.
xH2equiv
H2equiv 5
!0:333 xH2equiv 1xC1equiv
g
Richardson number; nðLBO Þ5 LBO (4)Chamberlain xH2equiv 5xH2 1xCM
Ds2 u2exp
  xC1equiv 5xC1 1xC2 1xC3 1xC4
qexp 0:5
Equivalent air source diameter; Ds 5dexp (8)
qair
(5)Chamberlain Wind Velocity at Flare Tip
Reference wind velocity, uwind0 is measured at a reference
Here, qair is ambient air density, and qexp is expanded source height, h0 (10 m). Wind velocity at the flare tip, uwind is
density. scaled according to flare stack height, where aw 5 power

Process Safety Progress (Vol.00, No.00) Published on behalf of the AIChE DOI 10.1002/prs Month 2016 5
exponent depends on ground roughness of surrounding
area: rural (0.14); built up rural (0.27); suburban (0.34). If m_ NG 5255:42d02 YNG
release elevation < 10 m, the 10 m value is used. ( 2
2:0074XNG 21:0103XNG 10:0185 ; for XNG > 0:665
  YNG 5
hstack aw 2:9013
uwind 5uwind;0 (9) 0:7873XNG ; for XNG  0:665
h0
LBO
XNG 5
VERTICAL FLAMES 174:83d00:8794
Vertical Flame Tilt (13)
Angle between vertical and flame buoyancy section axis,
a. Chamberlain predicts tilt for natural gas flames, a0
Vertical Flare Tip to Flame End Distance, LB
(
8000R ; for R < 0:05 This parameter is used by Chamberlain as an intermediate
a0 nðLBO Þ5 step in determining flame dimensions Lf and RL. Flame
1726ðR20:026Þ0:5 1134 ; for R  0:05 length decreases with increasing wind, due to air
(10)Chamberlain entrainment.
uwind
Flame velocity ratio; R5 (11)Chamberlain LB
uexp 50:4910:51e20:4uwind (14)Chamberlain
LBO
For gases other than natural gas, it is found that a correction
term is required since hydrogen tilts more than predicted by For tilt angles a > p/3 (608) approximately, it is found that
Chamberlain, and syngas less, see Figure 5. A correction can flame length actually increases and obtains a maximum
be applied based on the ratio of mass release rates (basically length of approximately LBO at a 5 p/2 (908 horizontal),
a momentum correction) using the concept of m_ NG 5 mass which is not predicted by Chamberlain, see Figure 6. The
release rate of natural gas that creates the same LBO length as Chamberlain model could not include the effect of flame
the gas of interest, for the same flare diameter. A correlation length increase for tilted flames with wind, since no data
for m_ NG has been developed through multiple runs of the were available at that time. An increase in flame length with
chamberlain model. This method successfully corrects both tilt has previously been observed, for example, by Johnson
hydrogen underprediction and syngas overprediction. et al. [13] for horizontal natural gas flames. Consequently, a
modification is required, which increases LB from its value at
m_ NG 608 tilt (LB60) to LBO at 908 tilt. This requires a several step
a5a0 (12)
m_ process, using R60, the value of R (5 uwind/uexp) that yields
Where, a tilt of 608:

LB
5ae ba when a < p=3
LBO
a5e 290b  
1 1
b5 ln
30 ðLB =LBO Þ60
ðLB =LBO Þ60 50:4910:51e20:4uwind;60
(15)
uwind;60 5R60 uexp
8   2
>
> m_ 60m_ nðLBO Þ
>
< 0:0261 60nðLBO Þ m_ NG 2134 =1726 ;
> for
m_ NG 8000
> 0:05
R60 5
>
> 60m_ nðLBO Þ
>
> 60m_ nðLBO Þ
: ; for  0:05
m_ NG 8000 m_ NG 8000

Vertical Flare Momentum Section Length, bl Vertical Flare Wind–Buoyancy Section Length, RL
The Chamberlain method is used directly in the new By geometry,
model since it found to be adequate for hydrogen and
syngas  0:5
RL 5 L2B 2ðblsin ðaÞÞ2 2blcos ðaÞ (17)Chamberlain
8
>
> 0:2LB ; for a50
>
>
>
>
< sin ðK aÞ
bl5 LB ; for a 6¼ 0; a 6¼ p K 50:185e220R 10:015
>
> sin ðaÞ Vertical Flare Centerline Flame Length, Lf
>
>
>
> By definition,
:
0:015LB ; for a5p
(16)Chamberlain Lf 5bl1RL (18)

6 Month 2016 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.00, No.00)
Figure 5. Vertical flame tilt vs. R for hydrogen and syngas. [Color figure can be viewed at wileyonlinelibrary.com]

trends toward 1 as Ly/LBO trends toward 0: a flame that stays


horizontal. Low jet momentum drives Richardson number up
and bl/LBO trends toward 0 as Ly/LBO trends toward 1: a
flame that rapidly becomes vertical.
The horizontal flame correlation does not include the
impact of wind (no available data) and so it is mathematical-
ly possible for a horizontal release to become more vertical
than the same flame if released vertically (where AP Flame
does account for wind), which is clearly wrong. To avoid
this, the maximum value for d is capped so that it cannot
exceed (p/2 2 a) where a is the tilt from vertical, for the
same flame released vertically. Lift angle, d is obtained by
simple geometry:
      
Ly p
d5MAX 0; MIN ASIN ; 2a (22)
Lf 2b 2
Figure 6. Vertical flame length as a function of tilt. [Color
figure can be viewed at wileyonlinelibrary.com] This means that each time a horizontal release is run in AP
Flame, an equivalent vertical release must also be run to
determine a. Note that if d is limited to (p/2 2 a), then Lf
HORIZONTAL FLAMES must be adjusted to match, Ly 5 (Lf 2 bl) sin (d).

Horizontal Flame Length, Lf 458 RELEASE FLAMES


Lf is found to be independent of wind within range of data
obtained and can be approximated to be LBO, see Figure 7a. Flame Length
Lf 5LBO (19) Based on Air Products R&D tests, the length of 458 release
flames is found to be similar to horizontal flames at the same
Horizontal Flame Momentum Section Length, bl release rate:
Momentum section length is found to be a function of
Richardson number, see Figure 7b. Switching LBO for Lf from Lf 5LBO (23)
(19) gives:
bl bl
520:125nðLBO Þ11:25; 0  1 (20) Flame Tilt
LBO LBO
Based on limited testing, 458 release flames can be
Horizontal Flame Lift, Ly and Lift Angle, d approximately modeled assuming that they are relatively
Horizontal flames lift in the wind–buoyancy section, due straight and continue at approximately a 458 angle from
to buoyancy. Lift is found to be a function of Richardson release point to the end of the flame.
number. See Figure 7c. Switching LBO for Lf from (19) gives:
Effect of Wind
Ly Ly No data are currently available to determine the impact of
50:125nðLBO Þ20:25; 0  1 (21) wind on 458 release flames. It would seem reasonable to
LBO LBO
assume that at high wind speeds the flame would bend
Equations (20) and (21) are mutually consistent. High jet toward horizontal, but no allowance has been made for this
momentum drives Richardson number down and bl/LBO in the model.

Process Safety Progress (Vol.00, No.00) Published on behalf of the AIChE DOI 10.1002/prs Month 2016 7
Figure 7. Horizontal flame dimensions for nat. gas, H2, and mixtures of nat. gas, H2, N2, and CO2. [Color figure can be viewed
at wileyonlinelibrary.com]

FLAME RADIANT HEAT FRACTION, F ground underneath the flame preheating some of the air
Total heat radiating from flame, Qrad is determined from entering the flame, increasing the flame temperature—this
F 5 radiant heat fraction, m_ 5 mass flow rate (kg/s), effect would diminish with distance between flame and
DHcombustion 5 gas heat of combustion (Lower Heating Value, ground and would be a lesser effect for vertical release
LHV) using Eq. (24). flames. It is not clear why this effect is not seen for natu-
1. AP Flame uses the Chamberlain correlation ral gas. Where the flame is very near the ground, it will
F 5 0.11 1 0.21 exp (20.00323 uexp) for vertical natural impinge on the ground and AP Flame is no longer really
gas flames. Data indicate that this correlation is also rea- valid and a cap of 10 is applied to Lf/hRC.
sonable for horizontal natural gas flames, see Figure 8a. 4. Radiant fractions are found to be similar for 458 and hori-
2. Figure 8b shows radiant fractions for all available hydrogen zontal flames at the same flow rates and example for
data, indicating that vertical release values of F can be well hydrogen/natural gas mixtures is shown in Figure 8d.
correlated with Mflux mass flux at pipe exit (not expanded Note that for this data F values are similar for natural gas
source). This figure also shows that equivalent horizontal and hydrogen, although this is not generally the case.
and 458 release hydrogen flames have higher F values. 5. Addition of inert gases reduces radiant fraction and a simple
3. To predict F values for horizontal and 458 release hydro- relationship where radiant fraction is proportional to the fuel
gen flames, a correction factor to the predicted vertical F fraction in the gas does a reasonable job capturing the effect
value (from the Mflux correlation) is applied, based on the within the range tested. Figure 8e summarizes Air Products
ratio of flame length, Lf, to elevation of radiant center R&D test results which were obtained for 458 flames, which
above ground, hRC, see Figure 8c. The correction factor proved to be more stable than horizontal flames at this scale.
increases as the distance between the flame radiant center The fuel flow was kept constant and inert added as required.
and the ground, hRC decreases. It can be speculated that It was not possible to obtain readings for higher inert addi-
the reason the F value is higher for a horizontal or 458 tion due to flame instability/blow-off.
flame, compared to an equivalent vertical flame, is due to 6. For purposes of the overall correlation for radiant frac-
reflection of thermal radiation off the ground between tion, the syngas composition is split up into two effective
the flame and the point of measurement and/or the hot components: C1equiv which is the total mole fraction of

8 Month 2016 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.00, No.00)
Figure 8. Radiant fraction. [Color figure can be viewed at wileyonlinelibrary.com]

“sooting” components C1 through C4 (assumes C2, C3,


C4 are only minor components in gas), H2equiv, which is
the total mole fraction of “nonsooting” components H2,
and CO, no data available on CO flames, so assumed
similar to H2 based on similar stoichiometry and flame
temperature. Note that C1equiv 1 H2equiv add up to less
than 1 when inerts are present, reflecting the “dilution”
effect of inerts. FC1 and FH2 are the radiant fractions of
“sooting” and “nonsooting” components, respectively.
7. The minimum values for F will occur for small high pres-
sure release flames, lowest value reported is 0.05 [14],
which acts as a lower limit in AP Flame.

_
Qrad 5F mDHcombustion

F 5MAXfðC1equiv FC1 1H2equiv FH2 Þ; 0:05g Figure 9. Distribution of radiant heat sources along flame
C1equiv 5xC1 1xC2 1xC3 1xC4 centerline. [Color figure can be viewed at wileyonlineli-
brary.com]
H2equiv 5xH2 1xCO
FC1 50:1110:21e 20:00323uexp
(
FH2 5Fcorr ð0:169120:01ln ðMflux ÞÞ hstack 1MAXð0; ð0:66Lf 2bÞsin ðdÞÞ ; for horizontal
hRC 5
4m_ hstack 10:66Lf sin ðp=4Þ ; for 45o
Mflux 5
pd02 (25)
8
<1; for vertical
Fcorr 5 DISTRIBUTION OF RADIATING SOURCES IN FLAME
: The model splits the total radiated heat up into 30 point
1:36 1 0:076 MINf10; Lf =hRC g ; for horizontal=45o
sources located evenly along the flame centerline (length Lf),
(24)Chamberlain following method by Ekoto [5]. Each point source, i, has a

Process Safety Progress (Vol.00, No.00) Published on behalf of the AIChE DOI 10.1002/prs Month 2016 9
Figure 10. Overall comparison of model predictions vs. data for some key flame parameters, Lf (m) and F. [Color figure can
be viewed at wileyonlinelibrary.com]

discrete location (xi, yi, zi) following along the flame center- relative humidity) and distance traveled through air (distance
line path of bl and RL. The amount of heat radiating from between radiation source and receiver). Method from Wayne
each point source has a weighting factor, wi arranged to be a [18] si 5 Air transmissivity 5 fraction of radiated heat passing
maximum at the radiant center of the flame, 2/3 along Lf (5 through distance Ri of air, Ri 5 distance from source i to
bl 1 RL) and a minimum at each end. Figure 9 shows how receiver, RH 5 Ambient relative humidity (fraction) Smm 5 sa-
the radiating heat is distributed. turated water pressure, mm Hg at ambient air temperature,
The radiating heat weighting factor for ith point in flame, Tair. It should be noted that the Wayne correlation was devel-
wi, is defined as, oped for hydrocarbon flames with both water and carbon
dioxide combustion gases, hydrogen only has water. Some
(
0:003273i ; for 1  i  20 inaccuracy is therefore expected using this correlation for
wi 5 (26) hydrogen. However, test results indicate this error to be rela-
0:003273ð2021:9ði220ÞÞ ; for 20  i  30 tively minor. For example, the estimation of radiation fraction
F for the Air Products full scale hydrogen flare test [4]
Where Qrad,i is the Heat radiation from the ith point source, obtained effectively the same calculated value for F (using
AP Flame) from two radiometer locations, one at 34 m from
Qrad;i 5Qrad wi (27) the flare tip, the other at 90 m.
    2
si 51:00620:01171log10 Xair ðWAÞi 20:02368 log10 Xair ðWAÞi
AIR TRANSMISSIVITY     2
Water and carbon dioxide in the air adsorb thermal radia- 20:03188 log10 Xair ðCDÞi 10:001164 log 10 Xair ðCDÞi
tion as a function of the concentration of carbon dioxide in (28)
the air (fixed), water in the air (function temperature and

10 Month 2016 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.00, No.00)
Table 2. Johnson et al. horizontal natural gas jet fire tests.

Test 1040 Test 1083 Test 1033 Test 1089


Data from Paper
Mass flow rate (kg/s) 2.8 8.4 7.9 3.8
Stagnation pressure (barg) 0.3 2.1 11.1 66
Stagnation temperature (K) 277 267 279 281
Hole diameter (mm) 152 152 75 20
Wind speed (m/s) 1.7 0.3 3.9 6.9
Wind direction, clockwise from north (8) 247 326 271 269
Ambient temperature (K) 279 281 282 286
RH (%) 89 80 81 91
Calculated Values
Wind to release angle (8) 223 56 1 21
AP Flame expanded source Mach number 0.49 1.25 1.9 2.07
AP Flame radiant fraction, F 0.22 0.16 0.14 0.14

Table 3. Johnson et al. heat fluxes vs. AP Flame predictions.

x y z DATA AP Flame x y z DATA AP Flame


Test 1040 (m) (m) (m) (kW/m2) (kW/m2) Ratio Test 1083 (m) (m) (m) (kW/m2) (kW/m2) Ratio
1 15 20.5 10.3 8.6 15.7 1.8 1 9 22 10.3 14 21 1.5
2 15 20.5 14.3 6.7 9 1.3 2 9 22 14.3 11.4 13.7 1.2
4 15 20.5 24.3 3.3 3.3 1.0 3 9 22 18.3 8 8.4 1.1
5 15 20.5 30.3 2.2 2.2 1.0 4 9 22 22.3 6.6 6.9 1.0
10 10 20.5 18.3 5.5 5.6 1.0 5 9 22 26.3 5.2 5.2 1.0
12 15 20.5 214.3 9.6 9 0.9 6 9 22 30.3 3.4 3.9 1.1
13 15 20.5 224.6 3.9 3.3 0.8 10 9 21 44.3 0.7 2 2.9
Average 1.1 12 50 22 0.3 4.6 5.7 1.2
13 55 22 0.3 3.3 3.7 1.1
14 60 22 0.3 2.2 3 1.4
Average 1.4

x y z DATA AP Flame x y z DATA AP Flame


Test 1033 (m) (m) (m) (kW/m2) (kW/m2) Ratio Test 1089 (m) (m) (m) (kW/m2) (kW/m2) Ratio

1 15 22 10.3 20.2 24.7 1.2 1 15 22 10.3 9.5 13.6 1.4


2 15 22 14.3 14.1 14.6 1.0 2 15 22 14.3 5.8 7.7 1.3
4 15 22 24.3 5.9 5.7 1.0 3 15 22 18.3 3.8 4.8 1.3
5 15 22 30.3 4 3.7 0.9 4 15 22 22.3 2.6 3.3 1.3
9 5 22 18.3 6.8 7.2 1.1 5 15 22 26.3 2 2.4 1.2
10 10 22 18.3 8.3 8.5 1.0 Average 1.3
12 15 22 214.3 14 14.6 1.0
13 15 22 224.6 5.7 5.6 1.0
Average 1.0

x, release direction; y, vertical; z, perpendicular to release direction.

Rh Ri Smm X 30
Qrad;i si
Xair ðWAÞi 5288:651 q5 (31)
T 4pRi2
(29) i51
Ri
Xair ðCDÞi 5273
 T  Where Ri is the distance from ith point in flame to receiver,
3816:42
23:189862
Smm 50:0075e Tair 246:13
(30)  0:5
Ri 5 ðx2xi Þ2 1ðy2yi Þ2 1ðz2zi Þ2 (32)

HEAT FLUX AT POINT OF INTEREST RECEIVER


OVERALL MODEL PREDICTIONS FOR FLAME LENGTH AND RADIANT
The heat total heat flux from the flame, at a point of inter- FRACTION VERSUS ALL AVAILABLE DATA
est receiver, q, kw/m2 is the sum of heat fluxes received Figure 10 shows overall comparisons of model vs. data
from each radiating point (i) in the flame. (x,y,z) is location for some key parameters of vertical, horizontal and 458
of receiver, (xi, yi, zi) is location of ith radiating source. flames. Overall, the model performs well. It is encouraging,

Process Safety Progress (Vol.00, No.00) Published on behalf of the AIChE DOI 10.1002/prs Month 2016 11
Figure 11. Johnson et al. test result flame profile side view vs. AP Flame prediction. [Color figure can be viewed at wileyonli-
nelibrary.com]

for example, to see the good prediction of the full scale same direction as the wind. For Test 1089 (high pressure
Garyville vertical syngas flame radiant fraction (0.12 model release), AP Flame overpredicts heat fluxes. For natural
vs. 0.11 data) and flame length (25 m model vs. 25 m data) gas, AP Flame simply uses the Chamberlain correlation for
using the method developed from small scale R&D tests on F and it is recognized in the Chamberlain paper that this
horizontal syngas mixture flames. correlation may be conservative for high pressure/high
velocity releases, 600 or less in diameter.
SUMMARY AND CONCLUSIONS
1. The new AP Flame model is based on the established NOMENCLATURE
Chamberlain model which has proven to be a strong foun- Symbol Description
dation. The new model extends the scope to include non- aw Power exponent for scaling wind at different
sooting gases such as hydrogen and carbon monoxide as heights
well as inerts and it also covers horizontal and 458 release a Parameter in vertical flame length correction
flames. b Parameter in vertical flame length correction
2. The model is based on a reasonably extensive dataset of bl Length of vertical flame momentum dominated
previously published and new test data on natural gas and (m)
hydrogen and more limited data on syngas mixtures. Data CDequiv Weighted carbon dioxide equivalent for CD, N2,
cover a wide range of scales from lab and pilot scale up WA components
to and including large industrial flares (0.200 to 4200 pipe C1equiv Sum C1, C2, C3, C4 mole fractions
diameters) and low pressure up to high pressure releases d Diameter (m)
(nonchoked and choked). d0 Flare exit diameter (m)
3. Given the inherent variability in the nature of flames and Ds Equivalent air source diameter (m)
the many challenges involved with setting up tests and F Radiant fraction of flame
measuring needed parameters, the new AP Flame per- Fs Stoichiometric mass fraction fuel in fuel 1 air
forms quite well across the full range of flames studied. g Gravitational constant, 9.81 (m/s2)
H Enthalpy (J/kg)
AP FLAME PREDICTIONS VERSUS JOHNSON ET AL.’S DATA DHcomb Gas heat of combustion (LHV; KJ/kg)
1. Johnson et al. [13] present a model for predicting flame H2equiv Sum H2, CO mol fractions
shape and thermal radiation for large natural gas jet fires, hstack Height of flare tip above grade (m)
based on testing at Spadeadam, UK in the late 1980s. h0 Reference height, 10 m (m)
These data were not used in the development of AP Flame hRC Height of flame radiant center above ground (m)
and can, therefore, be used as an independent check of m_ Release rate (kg/s)
the model. The paper includes the results for four sets of mw molecular weight of released gas
tests, summarized in Table 2. K Variable used to determine b
2. Flame shape and heat flux values, compared to AP Flame L Dimension of interest, such as flame length (m)
predictions are presented in Table 3 and Figure 11. LB Distance from flare tip to flame end (m)
3. Overall, AP Flame performs quite well, allowing that for LBO Flame length in no wind (m)
the first two tests, release direction is different than wind Lmb Mean beam length (m)
direction—AP Flame assumes release direction is in the Lf Centerline flame length (m)

12 Month 2016 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.00, No.00)
Ly Horizontal flame lift (m) f (L) Richardson number
P Pressure (Pa) T Air transmissivity
q Heat flux (kw/m2) Subscripts
Qrad Total heat radiated from flame (kw) ad Adiabatic flame condition
R Ratio wind velocity/expanded source velocity ad fuel Adiabatic flame condition for fuel components
RH Relative humidity only (no inerts)
RH2 Ratio H2 to total fuel fraction of gas, used in adi- air Ambient air condition
abatic flame temperature correlation atmos Atmospheric condition
Ri Distance from radiating source to receiver (m) C1 Methane
RL Length of flame buoyancy/wind dominated (m) C2 Ethane
RWA Ratio water/(water 1 carbon dioxide) C3 Propane
Smm Ambient air saturated water pressure (mm Hg) C4 Butane
T Temperature (K) H2 Hydrogen
u Velocity (m/s) CM Carbon monoxide
X Ratio of LBO to LBO at given Mach number CD Carbon dioxide
x Mol fraction choke Choked flow condition
x Distance from flare tip in wind direction (m) exp Expanded (effective) source condition
Xair (CD) Variable representing impact of CD in air on s flame Flame condition
Xair (WA) Variable representing impact of WA in air on s i ith radiating source in flame
Y Ratio of release rate/release rate at given Mach NG Natural gas
number N2 Nitrogen
y Vertical distance above grade (m) O2 Oxygen
YL Horizontal flame lift due to buoyancy (m) products Combustion products condition
z Cross wind distance (m) throat Release exit condition
a Vertical flame tilt from vertical due to wind WA Water
(radians) wind Wind condition at flare tip
a0 Vertical flame tilt from vertical due to wind for wind0 Wind condition at reference height (10m)
natural gas (radians) 0 Initial condition
b Density correction factor 60 Condition at 608 tilt
d Horizontal flame lift angle (radians)
q Density (kg/m3)
FULL DATASET

h T mdot Ta Wind m/s Lf b a Ly


Source H2 C1 CO CO2 N2 Direction (m) d00 (K) (kg/s) (K) @ 10 m F (m) (m) (8) (m)
Schefer [16] 1 0 0 0 0 V 5 0.2 288 0.36 288 0 0.07 0
Schefer [16] 1 0 0 0 0 V 5 0.3 288 0.06 288 0 0.08 0
Willoughby [10] 1 0 0 0 0 V 5.5 0.75 272 0.08 272 1.6 0.11 4.6 5
Willoughby [10] 1 0 0 0 0 V 5.5 0.75 275 0.2 275 1.9 0.13 6.4 5
Willoughby [10] 1 0 0 0 0 V 5.5 0.75 271 0.35 271 4.5 0.11 6.4 12
Willoughby [10] 1 0 0 0 0 V 5.5 2 275 0.08 275 2.5 0.14 5.5 15
Willoughby [10] 1 0 0 0 0 V 5.5 2 275 0.2 275 4.2 0.15 6.1 15
Willoughby [10] 1 0 0 0 0 V 5.5 2 273 0.4 273 3.2 0.13 7.3 12
Willoughby [10] 1 0 0 0 0 V 10 0.75 273 0.35 273 10 7 16
Willoughby [10] 1 0 0 0 0 V 10 0.75 273 0.04 273 10 3 60
Willoughby [10] 1 0 0 0 0 V 10 2 273 0.4 273 10 7 33
Willoughby [10] 1 0 0 0 0 V 10 2 273 0.04 273 10 6 81
AP/DNV GL [15] 1 0 0 0 0 H 3.25 0.8 279 1 279 3.4 0.15 17.4 12.1 3.5
AP/DNV GL [15] 1 0 0 0 0 H 3.25 2.1 279 7.5 279 2.6 0.20 48.5 30.8 11
AP/DNV GL [15] 1 0 0 0 0 H 3.25 2.1 279 6.6 279 2.6 0.20 44.6 28.4 9.8
DNV GL [9] 0 1 0 0 0 H 3.25 0.8 290 2.9 290 5 0.12 19.8 11.9 2.4
DNV GL [9] 0 1 0 0 0 H 3.25 1.4 290 9.6 290 5 0.14 37.8 23.5 6.4
DNV GL [9] 0 1 0 0 0 H 3.25 2 290 19.5 290 5 0.17 49.9 30.8 15
DNV GL [9] 0.24 0.76 0 0 0 H 3.25 0.8 290 2.7 290 5 0.12 17.6 12.4 2.8
DNV GL [9] 0.24 0.76 0 0 0 H 3.25 1.4 290 7.2 290 5 0.14 30.7 20.6 6.4
DNV GL [9] 0.24 0.76 0 0 0 H 3.25 2 290 16.9 290 5 0.16 45.2 28.5 12
Fishburne [6] 1 0 0 0 0 V 10 6 125 1.9 290 1.5 0.11
Fishburne [6] 1 0 0 0 0 V 10 10 125 17 290 1.5 0.10
Fishburne [6] 1 0 0 0 0 V 10 27 125 1.9 290 1.5 0.15
Fishburne [6] 1 0 0 0 0 V 10 27 125 17 290 1.5 0.12
Fishburne [6] 1 0 0 0 0 V 10 27 125 43 290 1.5 0.11
Fishburne [6] 1 0 0 0 0 V 10 27 250 43 290 1.5 0.11
Fishburne [6] 1 0 0 0 0 V 10 27 125 84 290 1.5 0.10
Fishburne [6] 1 0 0 0 0 V 10 27 125 126 290 1.5 0.09
Fishburne [6] 1 0 0 0 0 V 10 27 125 5.5 290 1.5 48

Process Safety Progress (Vol.00, No.00) Published on behalf of the AIChE DOI 10.1002/prs Month 2016 13
h T mdot Ta Wind m/s Lf b a Ly
Source H2 C1 CO CO2 N2 Direction (m) d00 (K) (kg/s) (K) @ 10 m F (m) (m) (8) (m)
Fishburne [6] 1 0 0 0 0 V 10 27 125 9.1 290 1.5 55
Fishburne [6] 1 0 0 0 0 V 10 27 125 22 290 1.5 110
Fishburne [6] 1 0 0 0 0 V 10 27 125 26 290 1.5 93
AP Garyville [4] 1 0 0 0 0 V 53 27 311 1.74 290 2.3 0.16 26 50
AP Garyville [4] 0.73 0.07 0.04 0.16 0 V 53 27 311 13 290 2 0.11 25 9
AP Pasadena [4] 1 0 0 0 0 V 30 36 310 1.3 290 9 33 80
AP R&D 1 0 0 0 0 458 1.5 1 290 0.005 290 2 0.24
AP R&D 1 0 0 0 0 H 3 1 290 0.005 290 2 0.20
AP R&D 0 1 0 0 0 458 1.5 1 290 0.027 290 2 0.21
AP R&D 0 1 0 0 0 H 3 1 290 0.027 290 2 0.24
AP R&D 0.21 0.79 0 0 0 458 1.5 1 290 0.025 290 2 0.26
AP R&D 0.5 0.5 0 0 0 458 1.5 1 290 0.023 290 2 0.23
AP R&D 0.69 0.31 0 0 0 458 1.5 1 290 0.019 290 2 0.23
AP R&D 0.88 0 0 0 0.12 458 1.5 1 290 0.014 290 2 0.21
AP R&D 0.79 0 0 0 0.21 458 1.5 1 290 0.024 290 2 0.16
AP R&D 0.72 0 0 0 0.28 458 1.5 1 290 0.032 290 2 0.13
AP R&D 0.88 0 0 0.12 0 458 1.5 1 290 0.02 290 2 0.23
AP R&D 0.79 0 0 0.21 0 458 1.5 1 290 0.034 290 2 0.20
AP R&D 0.75 0 0 0.25 0 458 1.5 1 290 0.042 290 2 0.17
AP R&D 0 0.83 0 0 0.17 458 1.5 1 290 0.036 290 2 0.19
AP R&D 0 0.71 0 0 0.29 458 1.5 1 290 0.045 290 2 0.16
AP R&D 0 0.83 0 0.17 0 458 1.5 1 290 0.041 290 2 0.19
AP R&D 0 1 0 0 0 H 2.8 1 290 0.027 290 2 3.7 1.5 2.4
AP R&D 0 1 0 0 0 H 2.8 1 290 0.053 290 2 4.6 2.1 1.8
AP R&D 0 1 0 0 0 H 2.8 1 290 0.064 290 2 4.9 3.0 1.8
AP R&D 0 0.83 0 0 0.17 H 2.8 1 290 0.036 290 2 3.4 1.8 1.8
AP R&D 0 0.71 0 0 0.29 H 2.8 1 290 0.045 290 2 3.0 2.1 0.6
AP R&D 0 0.66 0 0 0.34 H 2.8 1 290 0.05 290 2 4.0 2.1 1.2
AP R&D 0 0.83 0 0.17 0 H 2.8 1 290 0.041 290 2 4.0 2.7 1.8
AP R&D 0 0.73 0 0.27 0 H 2.8 1 290 0.053 290 2 4.0 2.1 1.5
AP R&D 0.21 0.79 0 0 0 H 2.8 1 290 0.025 290 2 4.3 2.1 2.4
AP R&D 0.5 0.5 0 0 0 H 2.8 1 290 0.023 290 2 4.3 1.5 2.1
AP R&D 0.69 0.31 0 0 0 H 2.8 1 290 0.019 290 2 3.4 1.5 1.8
AP R&D 1 0 0 0 0 H 2.8 1 290 0.003 290 2 2.1 0.3 2.0
AP R&D 1 0 0 0 0 H 2.8 1 290 0.005 290 2 3.0 1.2 1.8
AP R&D 0.93 0 0 0.07 0 H 2.8 1 290 0.012 290 2 2.7 1.5 0.9
AP R&D 0.88 0 0 0.12 0 H 2.8 1 290 0.02 290 2 2.4 1.8 0.3
AP R&D 0.79 0 0 0.21 0 H 2.8 1 290 0.034 290 2 2.4 2.4 0.0
AP R&D 0.74 0 0 0.26 0 H 2.8 1 290 0.043 290 2 2.4 2.4 0.0
AP R&D 0.88 0 0 0 0.12 H 2.8 1 290 0.014 290 2 2.7 1.5 0.6
AP R&D 0.79 0 0 0 0.21 H 2.8 1 290 0.024 290 2 3.0 1.8 0.6
AP R&D 0.72 0 0 0 0.28 H 2.8 1 290 0.032 290 2 2.1 2.1 0.0
Chamberlain [2] 0 1 0 0 0 V 10 12 290 10.3 290 8 0.23
Chamberlain [2] 0 1 0 0 0 V 10 12 290 17.4 290 8 0.20
Chamberlain [2] 0 1 0 0 0 V 10 12 290 22.3 290 8 0.19
Chamberlain [2] 0 1 0 0 0 V 10 12 290 24.6 290 8 0.15
Chamberlain [2] 0 1 0 0 0 V 10 8 290 5.9 290 8 0.21
Chamberlain [2] 0 1 0 0 0 V 10 8 290 8 290 8 0.20
Chamberlain [2] 0 1 0 0 0 V 10 8 290 11.2 287 8 0.17
Chamberlain [2] 0 1 0 0 0 V 10 8 290 14.9 290 8 0.15
Chamberlain [2] 0 1 0 0 0 V 10 8 290 20.3 290 8 0.15
Chamberlain [2] 0 1 0 0 0 V 10 6 290 4.6 290 8 0.16
Chamberlain [2] 0 1 0 0 0 V 10 6 290 5.9 290 8 0.14
Chamberlain [2] 0 1 0 0 0 V 10 6 290 7.6 290 8 0.13
Chamberlain [2] 0 1 0 0 0 V 10 6 290 19.1 290 8 0.12
Chamberlain [2] 0 1 0 0 0 V 10 42 290 27.2 290 8 0.29
Chamberlain [2] 0 1 0 0 0 V 10 42 290 35.8 290 8 0.26
Chamberlain [2] 0 1 0 0 0 V 10 42 290 43.6 290 8 0.28
Chamberlain [2] 0 1 0 0 0 V 10 42 290 55.6 286 8 0.27 55.5 36
Chamberlain [2] 0 1 0 0 0 V 10 6 290 1.54 290 8 0.28
Chamberlain [2] 0 1 0 0 0 V 10 6 290 2.63 290 8 0.22
Chamberlain [2] 0 1 0 0 0 V 10 6 290 3.8 290 8 0.19
Chamberlain [2] 0 1 0 0 0 V 10 24 290 4.6 290 8 0.32

14 Month 2016 Published on behalf of the AIChE DOI 10.1002/prs Process Safety Progress (Vol.00, No.00)
h T mdot Ta Wind m/s Lf b a Ly
Source H2 C1 CO CO2 N2 Direction (m) d00 (K) (kg/s) (K) @ 10 m F (m) (m) (8) (m)
Chamberlain [2] 0 1 0 0 0 V 10 24 290 5.7 290 8 0.31
Chamberlain [2] 0 1 0 0 0 V 10 24 290 8.96 290 8 0.30
Chamberlain [2] 0 1 0 0 0 V 10 24 290 12.2 290 8 0.30
Chamberlain [2] 0 1 0 0 0 V 10 24 290 12.4 290 8 0.29
Chamberlain [2] 0 1 0 0 0 V 10 24 290 14.9 290 8 0.30
Shirvil [7] 1 0 0 0 0 H 1.5 0.2 290 0.1 290 2 5.8 5.5 0.2
Shirvil [7] 1 0 0 0 0 H 1.5 0.1 290 0.018 290 2 2.9 2.7 0.1
Shirvil [7] 1 0 0 0 0 H 1.5 0.1 290 0.035 290 2 3.6 3.5 0.1
Shirvil [7] 1 0 0 0 0 H 1.5 0.1 290 0.045 290 2 4 3.8 0.2

Notes
1. Chamberlain and Fishburne do not provide flare elevation, but do provide wind speed. For this data, a 10 m nominal eleva-
tion has been selected.
2. Willoughby radiant fractions estimated from averaging published heat fluxes at different locations.
3. All flame dimensions determined from photographs. Since flames can move significantly over time due to wind fluctuation,
values provided are subject to error. AP Garyville hydrogen and syngas flame dimensions obtained accounting for perspective,
according to location of camera vs. flame.
4. Chamberlain radiant fractions vs. release conditions obtained through interpretation of plots provided in paper.

10. D.B. Willoughby, M. Royle, S. Nilsen, and T. Gautier,


LITERATURE CITED Hydrogen Venting under Variable Flow Conditions,
1. D. Miller, S. Jung, and E. Lutostansky, “Applicability of Health and Safety Laboratory, UK, 2011.
currently available flare radiation models for hydrogen 11. K.B. Yuceil and M.V. Otugen, Scaling parameters for
and syngas,” Proceedings of 10th Global Congress on underexpanded supersonic jets, Phys Fluids 14 (2002),
Process Safety, New Orleans, LA, 2014. 4206.
2. G.A. Chamberlain, Developments in design methods for 12. B.J. McBride and S. Gordon, NASA CEA, Available at
predicting thermal radiation from flares, Chem Eng Res http://cearun.grc.nasa.gov/index.html.
Des 65 (1987), 299–309. 13. A.D. Johnson, H.M. Brightwell, and A.J. Carsley, “A mod-
3. Pressure Relieving and Depressurizing Systems, American el for predicting the thermal radiation hazards from large
Petroleum Institute (API) 521, 6th Edition, 2014. scale horizontal natural gas jet fires,” Hazards XII, UMIST,
4. D. Miller, and J. Bedenbaugh, “Hydrogen and syngas Manchester, UK, April 19–21, 1994.
flares – Comparison between experimental data and 14. R.W. Molina and R.W. Schefer, et al., “Radiative fraction
model predictions,” AFPRC Industrial Combustion Sym- and optical thickness in large – Scale hydrogen – Jet
posium, Houston TX, 2014. fires,” Proceedings of the Combustion Institute, 2007.
5. I.W. Ekoto, A.J. Ruggles, L.W. Creitz, J.X. Li, “Updated jet 15. Advantica, Report on Hydrogen Jet Fires Carried Out On
flame radiation modeling with buoyancy corrections,” Behalf of Air Products & Chemicals Inc., Private Report,
International Conference on Hydrogen Safety, Brussels, 2009.
16. B.C.R. Ewan and K. Moodie, Structure and velocity meas-
Belgium, 2013.
urements in underexpanded jets, Combust Sci Technol 45
6. E.S. Fishburne and H.S. Pergament, “The dynamics and
(1986), 275–288.
radiant intensity of large hydrogen flames,” Colloquium 17. A.D. Birch, D.J. Hughes, and F. Swaffield, Velocity decay
on Fire and Explosion, 1979. of high pressure jets, Combust Sci Technol 52 (1987),
7. L.C. Shirvill, P. Roberts, C.J. Butler, T.A. Roberts, and M. 161–171.
Royle, Characterisation of the Hazards from Jet Releases 18. F.D. Wayne, An economical formula for calculating atmo-
of Hydrogen, Health and Safety Laboratory, UK, 2005. spheric infrared transmissivities, J Loss Prev Process Ind 4
8. R.W. Schefer, W.G. Houf, B. Bourne, and J. Colton, Spa- (1991), 86–92.
tial and radiative properties of an open-flame hydrogen 19. A.D. Birch, D.R. Brown, M.G. Dodson, and F. Swaffield,
plume, Int J Hydrogen Energy 31 (2006), 1332–1340. The structure and concentration decay of high pressure
9. B.J. Lowesmith, An experimental programme to study jets of natural gas, Combust Sci Technol 36 (1984), 249–
high pressure jet fires following releases of natural gas/ 261.
hydrogen mixtures, R0051-WP2-R-0 (Part of Deliverable 20. G.T. Kalghatgi, Flowout stability of gaseous jet diffusion
D37), 2008. flames. Part 1: In still air, Combust Sci Technol (1981).

Process Safety Progress (Vol.00, No.00) Published on behalf of the AIChE DOI 10.1002/prs Month 2016 15