IRII 7457

• • •
......
-
Bureau of Mines Report of Investigations I December 1970

Hydrogen Flare Stack Diffusion Flames:

Low and High Flow Instabilities,
Burning Rates, Dilution Limits, ,
f...
Temperatures, and Wind Effects

N71-1210tj•
G ~RU)
(CATEGORY)

UNITED STATES DEPARTMENT OF THE INTERIOR

Report of Investigations 7457

Hydrogen Flare Stack Diffusion Flames:

Low and High Flow Instabilities,
Burning Rates, Dilution Limits,
Temperatures, and Wind Effects

By J. Grumer, A. Strasser, J. M. Singer, Patricia M. Gussey,

and Valeria R. Rowe

UNITED STATES DEPARTMENT OF THE INTERIOR

Walter J. Hickel, Secretary

BUREAU OF MINES
Elburt F. Osborn" Director

The work !Jpon which this report is based was done by the Bureau of Mines, U.s. Department of the Interior,
tIlder a working fund agreement with the Space Nuclear Propulsion Offl':!, Cleveland, CIIio.
 CONTENTS

Abstract................................................................. 1
Introduction.. . . . . . . . .. .. . . . . . . . .. . . .. . .. .. ... . . . . .. . .. ... ..... . .. .. ... . . 2
Low flow instability limit............................................... 2
High flow instability limit ..........................•................... 11
Dilution limits or blowout limits of highly diluted diffusion flames of
hydrogen-inert gas mixtures............................................ 15
Burning rates............................................................ 17
Temperatures of hydrogen diffusion flames •.•.•••••••••••••••••••••••••••• 19
Flame characteristics in winds........................................... 22
Stripping of hydrogen from its diffusion flame by winds ••••••••••••• 22
Blowback of hydrogen diffusion flames by winds .••••••••••••••••••••• 24
Blowout limits of hydrogen-nitrogen diffusion flamer. in crosswinds.. 31
SUIIIlla ry . . • • • • • • • • • • • • . • • • . • • . • • • • . • • . • • • • • • • . . . • • . • • • • • • .. • • • • • • • • • . • . . • . . 32
References. . • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

ILLUSTRATIONS

1. Twin 8-inch-id flare stack.......................................... 4

2. Dip limits of hydrogen diffusion flames into stack.................. 6
3. Experimental flare stacks showing sampling points.................. 8
4. Two views of a hydrogen diffusion flame on a l2-inch stack, flow
rate = 0.66 cfs..... .............................................. 10
5. Blowout limits and stable flame point of diffusion flames in air of
mixtures of hydrogen plus inert gases ••••••••••••••••••••••••••.•• 13
6. Lift-off of a hydrogen diffusion flame from a 0.77-inch-diameter
burner............................................................ 15
7. T~mperatures on the axis of a hydrogen diffusion flame •••.•••••••••• 20
8. Sampling equipment for hydrogen stripping by cross . .olinds.............
. 23
9. Average temperature observed inside a 4-inch-id stack burning at a
12 ft/sec hydrogen velocity and facing into air blast ••••••••••••• 25
10 . Average temperature observed inside a 4-inch-id stack burning at a
22 ft/sec hydrogen velocity and facing into air blast ••••••••••••• 26
11. Average temperature observed inside a 4-inch-id stack burning at a
38 ft/sec hydrogen velocity and facing into air blast ••••••••••••• 28
12. Effective hydrogen concentration in 4-inch-id stack uith diffusion
flame burning against opposed wind •••••••••••••••••••••••••••••.•• 29

TABLES

1. Field experienced ltmits and predicted limits of fires in hydrogen

flare stacks...................................................... 3
2. Flamedip and flameback limits for hydrogen diffusion flames in
open air............................... ........................... 7
3. Percentages of air inside of an J.8-inch-id stack with lpward helium
flow 0 f 0 . 5 0 r 0 . 8 c f s . . . . . . . . . . . . . . . . . III • • • • • • • • • • • • • • • • • • • • • • • • • • 7
4. Blowout and lift-off limits of diffusion flames of hydrogen-nitrogen
mixtures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
ii

TABLES--Continued

5. Blowout limits and stable flame data for diffusion flames on flare
stacks in air of mixtures of hydrogen plus inert gases •••••••••••• 14
6. Composition of gases near lifted diffusion flames of hydrogen-
nitrogen mixtures ••••••••••••••••••••••••••••••••••••••••••••••••• 15
7. Dilution limits of diffusion flames in air of mixtures of hydrogen
and nitrogen...................................................... 16
8. Burning rates of hydrogen diffusion flames •••••••••••••••••••••••••• 18
9. Predicted heights of hydrogen diffusion flames versus heights
obseI"V'ed in the field............................................. 18
10. Temperatures and analyses of gases on axis of a diffusion flame
burning 3.5 cfm of hyd~ogen on a 4-inch-id burner ••••••••••••••••• 21
11. Probe positions when sampling around hydrogen diffusion flames on a
vertical 4-inch-id stack.......................................... 23
12. Heights of hydrogen peaks on gas chromatographic charts. Analyses
of samples taken with integrating probes •••••••••••••••••••••••••• 24
13. Temperatures and analyses of gases inside of a horizontal 4-inch-id
diffusion flame burner, facing into an opposing wind. Hydrogen
velocity, 12 ft/sec; axial air velocity, 45 ft/sec •••••••••••••••• 27
14. Temperatures and analyses of gases inside of a horizontal 4-inch-id
diffusion flame burner, faCing into an opposing wind. Hydrogen
velocity, 38 ft/sec; axial air velocity, 42 ft/sec •••••••••••••••• 27
15. Total and static pressures and corresponding air velocities produced
by a wind from an 8-inch-id duct blowing against a coaxial 4-inch-
diameter horizontal stack ••••••••••••••••••••••••••••••••••••••••• 30
16. Effect of cros~wind on blowout limits of hydrogen-nitrogen diffusion
flames on a 4-inch-diameter stack ••••••••••••••••••••••••••••••••• 32
 HYDROGEN FLARE STACK DIFFUSION FLAMES: LOW AND HIGH FLOW
INSTABILITIES, BURNING RATES, DILUTION LIMITS,
TEMPERATURES, AND WIND EFFECTS

by

J. Grumer,l A. Strasser,2 J. M. Singer,3 Patricio M. Gussey,4 and Valerio R. Rowe 3

ABSTRACT

The Bureau of Mines, under the sponsorship of the Space Nuclear Propul-
sion Office, conducted a laboratory-scale hydrogen safety study which deter-
mined several combustion characteristics of hydrogen diffusion flames.

Experiments show that ambient air may enter the top of a hydrogen flare
stack when the hydrogen flow is low. A new concept, supported by photographic
evidence, predicts that diffusion flames burning in air on a wide, upright
pipe (stack) and fed with slow, upward flows of buoyant gas will induce a
downward flow of air along the walls of the pipe that can support combustion
within the pipe. Predicted f1amedip limits agree roughly with experimental
values determined on 6-, 12-, and 18-inch-diameter stacks and increase with
increasing stack diameter.

Measurements were made of the limiting flow at which a hydrogen diffusion

flame blows out in still air. By means of an empirical application of the
critic~l boundary velocity gradient concept, these data lead to a blowout
limit of about 108 reciprocal seconds for a hydrogen diffusion flame. Burning
rates of large hydrogen diffusion flames ranging from about 0.03 to 1 ft/sec
were used to predict approximate flame heights on flare stacks.

Temperatures of larger hydrogen diffusion flames up to about 3,600 0 F

were observed, but the most representative value appears to be about 2,600 0 F.

It was found that crosswinds do not strip significant amounts of unburned

hydrogen from its diffusion flame and that water-cooled flare stacks are not
likely to be damaged when flame is blo'WIl back into them by opposing winds.

lActing project coordinator and s llpervisory research chemist.

2Research physicist.
3Research chemist.
4 Chemist.
All authors are with the Safety Research Center, Bureau of Mines,
Pittsburgh, Pa l
2

INTRODUCTION

Although there have been accidents, industrial flare stack operations

have a good safety record. Usually gases being flared have low heating values
and are slow burning. Flow rates are rather steady and within design limits.
Most frequently, diffusion flames are employed, that is, no air is knowingly
mixed with the combustitle stream within the piping or in the flare stack
(burner). All air for combustion comes from the open atmosphere surrounding
the flame, downstream of the flare stack exit. This paper considers only
diffusion flames.

Flare stack operations required in the research and development of space

vehicles have also enjoyed a good safety record. In these operations, fast-
burning hydrogen is flared. Malfunctions of hydrogen flare stacks have gener-
ally occurred at low flows. Flows vary from the small flows due to boiloff
from storage vessels to flows of about 300 lb/sec for about a half hour. Thus
it may be necessary to flare more than 200 tons of hydrogen quickly and safely.
Although no blowout of a neat hydrogen diffusion flame has been re~orted yet,
the possibility of such blowout at high flows of hydrogen must be considered.
The problem of flame height on flare stacks, which is related to the burning
rate, and the problem of maintaining a hydrogen flame heavily diluted with
inerts such as helium, nitrogen, or steam have recently interjected themselves
into the planning of space test stands; concern has arisen over wind effects
and other factors.

LOW FLOW INSTABILITY LIMIT

At present when hydrogen is dumped at a test stand, it is either vented

unburned, burned as a diffusion flame on a flare stack (12),6 or vented over
a burn pond (22). In burn ponds, the hydrogen is dispersed through a pipe
manifold submerged in water and bubbled into the atmosphere where it is
I·
ignited and burned; this method is not considered here. Industry, particu-
larly the petrochemical industry, has realized the hazard of flame receding
into a line and of flow reversal during flaring of waste gas. It recognizes
the advisability of using a "continuous purge in flare systems where the aver-
age flow is too low to support stable combustion" (1). Such maleffects have
been attributed to air initially present in a flare system or entering through
openings in the lines. The present paper proposes a new mechanism explaining
malfunctions due to low flow.

Most previous investigators (1, 9, 12, 15) have treated the problem of
low flow instability as one involving premixed flames, that is, hydrogen and
air are assumed to mix inside the flare stack prior to burning. In fact, the
problem is more likely to be one of diffusional burning; air and fuel mix at
the flame surface as burning progresses. For example, Hajek and Ludwig (9)
recognized that all flare stacks which burn combustible mixtures have a st.able
operating velocity range outside of which they do not operate safely. They
also recognized that below the low velocity limit all or part of the flame may

6----~~~~--~--~------~------~----~----------~~----~-------------
Underlined numbers in parentheses refer to items in the list of references at
the end of this report.
 3

drop into the stack. They assumed that the flashback theory for premixed
flames on gas burners (I, B, 13, 24) was relevant to the low flow instability
limit of hydroger, rliffusion flames on flare stacks. Their treatment of such
flame instability was based on information by Lewis and Von Elbe (13) and
Von Elbe and Mentser (24), which treat the problems of flashback and flame
tilt of premixed flames. When flashback occurs the premixed flame moves
upstream into the burner until it is quenched or stabilized by a change in
channel geometry or local fuel-air composition. When flame tilt occurs, the
premixed flame partially enters the port; the plane of the flame base is at
some angle to the plane of the port. Flame tilt can degenerate into flashback
as the walls are wl,rmed by the tilted flame. Both of these flame phenomena
were treated by considering the balancing of counter current velocities;
namely, the burning velocity close to the wall and the stream velocity close
to the wall. Instability due to flame tilt (~), which is likely on lp.rge
diameter burners, was not considered by Hajek and Ludwig, nor-was the informa-
tion in reference B which indicates that classical flashback and tilted flame
theories are inadequate for predicting the low flow instability limit fer wide
burners. Moreover, Hajek and Ludwig make two assumptions that are not re~Bvn­
able. First, they assume that at a Reynolds number of 2100 the flashback
velocity is the same for laminar and turbulent flow. Secondly, they derive an
equation for a Reynolds number of 2100 but apply it generally. An empirical
coefficient is also employed.

In applying either the classical or the empirical flashback equations to

flashback limits of flare stacks, the percentage of hydrogen in the flare
stack gases must be given. The composition selected may correspond to the
rich flammability limit of hydrogen (74 percent hydrogen in air), or more
conservatively, to the peak flashback limit (36 percent hydrogen in air) (B).
Mass flows at the flashback limits of these two hydrogen-air mixtures have
been ca~culated for three flare stacks. The limit flows are given in table 1
and compared with the flows at which fire or no fire inside the stack has
been reported. Only the Hajek and Ludwig equation (calculated for 74 perc~nt
hydrogen i~ air) fits a field observation; it fits the observation made with
the R-inch twin-flare stack (fig. 1) but not with the lB- and 42-inch single
stacks. Thus, neither the Hajek-Ludwig nor the classical equation is relevant
to the hydrogen flare-stack low flow problem.

TABLE 1. - Field experienced limits and predicted limits of fires

in hydrogen flare stacks

Predicted h lb sec
Flare stack Actual hydrogen flows, Classical
diameter, lb sec flashback
in. Fire No fire l--...:e~u:.::aTt..::i.:.on:':---f~~;':':'oF=-~II.oC...-tTh is study
in stack in stJ!ck 36% Ha 74% H:a
in air in air
Twin B•••••••• 0.10 to 0.35 0.4 0.11 0.028 0.26
lB. • • • • • • • • • •• 7.5 X 10- 6 1 9 •1 X 10-" .32 .OB 1.1 X 10-2
42. • ~ • • • • • • • • • !-1....;4~.~9_X--..;:;1;.;;0_-3.....L.._>..:,•.=,3.=.2......&->..:,• .=,0.=.,8.......a.-..::-=-----L-.=..~__'___:..;.1;.,::2::..-__
1Manufacturer's specified Ilrlnilll.lm flow.
4

~
r---~~"~

\Pilot flame Pilot flame

Sin Bin

Common hydrogen
manifold

FIGURE 1. • Twin 8·lnch·ID Flare Stack.

Consider the flare stack in table 1 in which hydrogen was being flared
through two 8-inch-id twin stacks connected (as shown in fig. 1) to a common
manifold. For a flow of 0.10 to 0.35 lb/sec of hydrogen, flame formed inside
one of the flare stacks near the manifold and oscillated between the two
stacks. Both flare stacks operated satisfactorily when the total flow into
the manifold ",aG 0.4 lb/sec. Experiments showed that air can be inducted
through one of the stacks while buoyant gases flow up the other. When helium
flowed i nto the system, Pitot tube readings at the top of the flare stacks
were negative for one stack and positive for the other.

A mechanism may be advanced to explain flame instability at low flows of

hydrogen, employing the twin flare stack arrangement in figure I as an example.
It may be assumed that flow of air down one flare stack is due to the buoyancy
head of a column of hydrogen flowing in the other stack. The resistance to
the downward air flow is due to friction with the stack walls. (Frictional
pressure due to hydrogen flow is n~glected because it is balanced by the pres-
sure head causing hydrogen to flow.) The frictional head for turbulent flow
~P~, ft (l1), is given by the following equation:

(~Pa.)a1r = [ l6L Pa1r

• .2~..6 a1r
V-a1r Til)
[
0.0036 + 0.24(2/Re)· 36 air ]
(1)
 5

The buoyancy head ~P (23) is given by the following equation:

- gI.. p. 1 r (1- dH ) (2)

~

Equating I and 2 we get 3 which gives Va1r , ft 3 per sec, the maximum flow of
air which can be inducted by this system is as follows:

n[g (1-~2)Jl/2 O.lr 2 • 6

V. 1r - 4 [ 0.0036 + 0.24 (2 / Re).lr .36J 172
( 3)

L the length, ftj g, the gravitational constant, ft/sec 2 ; d, the specific

gravity, compared with air at I atmj p, the density, lb/ft3 j 0, the burner
diameter, ftj and Re, the Reynolds number.

If air and hydrogen mix in the flare stack carrying the hydrogen, and if
the resulting mixture contains 74 percent hydrogen, i~ ~s flammable. Pilot
flames on top of flare stacks can ignite the mixture, and flame may propagate
into the flare stack. After flame propagates down one stack, ~ir induction
down the other depends on the buoyancy of the combustion products flowing up
the first. Extensive flame propagation into the flare stack is impossible at
highe~ flows of hydrogen because the local concentrations of hydrogen are
likely to exceed the rich limit of flammability. A sharp limit between flame
dip and no flame dip is unlikely, because here and there a pocket of flammable
mixture may form. As the hydroge~ flow is increased, even this fragmentary
flame dip becomes unlikely. Thus the proposed mechanism provides a basis for
calculating the minimum flow of hydrogen above which flame is not expected to
penetrate into the flare stack. Equation 4 yields the minimum flow of
hydrogen,
vHe! - (0.74/0.26) V. 1r • (4)

The proposed concept yields results that are not in conflict with field
experience available to the authors. As shown in table 1, the predicted limit
for flame stability on the 8-inch twin-flare stack is 0.26 lb/st'c. Experience
shows the limit to be less than 0.4 lb/sec, more than 0.10 lb/sec, and perhaps
more than 0.35 lb/sec. Fire in an 18-inch-id single-flare stack has been
observed while hydrogen was flowing at about 7.5 X 10-6 lb/sec. The minimum
flow stated in the manufacturer's operating instructions for this flare stlck
is about I X 10.3 lb/sec. The concept proposed in this study leads to a mini-
mum hydrogen flow of about 1.1 X 10-2 lb/sec.
When a single stack is used, additional assumptions are necessary in
order to assign values to O.lr and Re. 1r in equation 3. In tbe twin-flare-
stack example, one flare stack carries the downflow of air, and the other
stack carries the upflow of hydrogen. In the single-stack case, one stack has
to carry both flows. Accordingly, two additional assumptions were made.
First, 74 percent of the cross-secticnal area of the stack was assigned to the
hydrogen flow and the remaining annulus of 26 percent to the countercurrent
air flow. This matches the earlier assumption that the hydrogen-air
6

combination of interest when mixed contains 74 percent hydrogen. Second, the

frictional pressure head due to the downward flow of air in the pipe was taken
to be equivalent to the frictional pressure head for a circular pipe whose
radius is the equivalent hydraulic radius of the annulus assumed to carry the
downward airflow.

The experimental points observed in this laboratory with the 6-, 12-,
and 18-inch stacks at which flame
was observed to begin to enter
each flare stack are presented in
figure 2 as f1amedip limits. No
sharp limits could be obtained, as
would be ~xpected. The obs~rved
limits agree roughly with the pre-
dicted limits for the same diame-
ter stacks. Equation 3 was modi-
fied for the prediction of the
10-1 value for the 6-inch stack by
assuming laminar flow. Turbulent
No flame dip flow was assumed for the 12- and
18-inch stacks because the
Reynolds numbers were near 2000,
and the stack lengths were four.
pipe diameters. It is unlikely
Flame dip that laminar flow will exist in
the field when large-diameter
flare stacks are employed.
Table 2 gives other characteris-
tics about these e:{periment~ and
also about f1ameback limits, which
Theoretical are the flows at which the flame
curve had moved into the stack so that
D~st of it was inside the pipe.

Further evidence that air can

flow downward from the open atmos-
phere into a stack was obtained by
experiments using helium in place
of hydrogen (table 3). Samples
were taken inside of the lS-inch-
id stack through which helium was
flo\dng at the rate of 0.5 cfs.
The experimental arrangem~!t is
shown in figure lA. With a loosely
fitting cover on the stack, helium
percentages were about 90 to 100.
With the cover removed and the
saMe flow maintained, percentages
of air were 85 to 70 1 inch inside
FIGURE 2•• Dip limits of Hydrogen Diffusion stack and <70 about 2 feet inside
Flanws Into Stack. of the stack. \See table 3,
 7

part A.) The experiments with helium clearly indicate that had low flows of
hydrogen been used, air would have flowed into the stack. If the hydrogen had
been ignited, it could have burned inside of the stack, although no air was
sent into the system upstream of the stack.

TABLE 2. - Flamedip and flameback limits for hydrogen

diffusion flames in open air

F1amedi limits Flameback

Stack diameter, Predicted, Experimental, Velocity, Reynolds limit
inches Ib/sec lb/sec ft/sec No.

6 ••••••••••••• 1.3 X Hi 3.5 X 1(i 0.36 170

12 •...•..•..... 2.9 X 10- 3
4.7 X 10- 3 1 •
1 20 11,100
18 ...•.•.•..... 1.1 X 10- 2 6.9 X la- 3 1 .78 1 1 ,100
42 ••••••••••••• .12 2.4 7,900
52 .............. .18 2.5 10 000
lExperimental flamedip values, based on hydrogen flow through entire cross
section. Others predicted.
TABLE 3. - Percerltages of air inside of an l8-ineh-id stack
with upward helium flow of 0.5 or 0.8 cfs

Flare stack ~~~~~~s~i~t~io~n~cTo~o~r~d~i~n~a~te~s~~i~n~c~h~e~s~Air, percent

From to From wall
WITH LOOSE COVER ON SJ:ACK
Apparatus 3A ••• 8 1 12
8 9 o
8 9 10
Apparatus 3B ••• 136 9 29
136 9 5
WITH STACK opm TO ATMOSPHERE
Apparatus 3A ••• 1 1 85
1 5 69
1 9 69
4 1 79
4 5 80
8 1 75
8 1 74
8 9 69
12 1 68
12 9 68
18 1 70
18 9 65
24 9 66
Apparatus 3B ••• 361 6 270
3 61 6 59
436 9 271
436 9 66
Sampling point A. Sampling point B.
20.8 efs. ·Sampling point C.
8

Four helium sampling

I probes shown,
40in copper tubing

.-11R..in-tti."m•• t . pipe
.-in-thick wall
6ft long
in

A,B. C,
Sampling point

Gasket
material
Pebble bed
approximately
6in deep

A Without inverted cop B With inverted cop

FIGURE 3. - Experimental Flare Stacks Showing Sampling Points.

An assembly consisting of an inverted cap over an upright pipe located

deep within a flare stack is called a molecular seal (18). It is often used
to prevent air from entering from the top of the flare stack. According to
Reed (18) "a commercially available molecular seal installed iDlDediately below
the flare at the top of the riser will establish perfect safety from entry of
air to the flRre system when the purge volume admitted is capabl~ of a line
velocity of frC'm 0.10 to 0.15 fo~t per second ••• " The top of the riser, that
is the top of th~ inverted cap, may be ruany feet below the top of the flare
stack. For example, Lapin (12) in his tests of hydrogen flare stacks used a
commercially available molecular seal, which was about 12 feet below the top
of the flare stack. ~H~se inverted cap systems are placed far enough inside
of the flare stack to keep them away from the high temperatures at the top of
the flare section (12).

It does not appear reasonable in view of the foregoing discussion that a

trap of this sort would prevent all ingress of air into a flare stack system
 9

against a slow flow of buoyant gas. Air should be able to fall to the down-
stream exit of the trap, as demonstrated by the following experiments: A
simulated molecular seal was constructed to fit the l8-inch-diameter stack
(fig. 3B). With a loose cover on the stack, helium at a rate of about 0.8 cfs
was used to flush air out of the stack and the seal. This flow corresponds to
a linear velocity of 0.43 ft/sec in the l8-inch-diameter section and 0.78
ft/sec in the annulus formed by the l-foot-diameter seal. In two minutes all
the air was flushed out of the stack. The cover. was removed with the helium
continuing to flow. After 1 minute, samples were taken ae poinls A, B, ~nd C
in figure 3B. Sampling time was approximately 1 minute. Point A is located
inside the seal just below its top, point B is at the bottom of the seal at
its downstream end, and point C is immediately above the seal on the axis of
the stack. A second run was made sampling at the same points but with the
flow reduced to about 0.5 cfs, which corresponds to 0.30 ft/sec in the stack
and 0.55 ft/sec in the annulus. Helium flow was continued for 3 minutes with
the cover on. Sampling was started 1 minute after the cover was removed with
the flow continuing. As shown by the data in part B of table 3, air entered
the stack when the cover was removed. At point B the percentage of air was
70 in the first run and 59 in the second. Thus the seal was not affective
here. Point C, where the air concentration was 66 percent, was 36 inches down
the stack. At point A within the seal the observed air percentages were 9 and
5, respectively. Clearly ingress of air into the stack was not arrested; only
the ingress of air into the seal was arrested.

Motion pictures were taken of flames on the l2-inch stack at the flamedip
limit and at lower flow rates, dropping progressively almost to the flameback
limit. Flames on the 18-inch stack at about the flamedip iimit were also
filmed. A frame-by-frame analysis of tnese films show~d that the flame
heights fluctuated considerably, at a frequency of about 2 to 3 fluctuations
per second for the l2-inch stack and about 2 fluctuations per second for the
l8-inch stack. An example of the extremes observed in flame shape and height
-, is given in figure 4 which shows two 0.04-sec views of a 0.66 cfs (3.3 X 10-3
lb/sec) hydrogen flame on the l2-inch stack. This flow (0.66 cfs) is below
the experimental flame dip limit but above the experimental. flameback limit.
The flame in the right hand frame is almost twice the height of the other.
More significantly, large voids exist at the base and within the body of the
flame3. These voids moved around and are attributed to flame quenching by the
combustion products accumulated in the immediate vicinity. As these pockets
of burned gas are dissipated by diffusion and slow convection, air comes in
and '~eals" the flame void. In the meantime, a pocket resulting in flame
extinguishment grows elsewhere. At higher flow rates, burned gases are
removed and air is entrained much faster, so that burning becomes continuous
up to extremely high rates of flow which may again disrupt the turbulent
flame (~, l!). In connection with the occurrence of holes in diffusion
flames, it may be noted that surprisingly high concentrations of nitrogen have
been obs~rved underneath hydrocarbon diffusion flames (6, lQ-2l) , indicating
that considerable convective transport occurs across diffusion flame surfaces
or through gaps in flame as shown in figure 4.

The exposure time of the motion pictures was about 0.0167 sec/frame.
St.ills of methane diffusion flames taken at the same speed also showed voids
10

FI GURE 4. - Two Views of a Hydrogen Diffusion F!ame on a 12-lnch Stack,

F low Rate"" 0.66 cfs.

in the flame, indicating that such voids are not limited to hydrogen nor
probably to wide stacks.

Experiments were also conducted with low-flow-hydrogen flames on the

l2-inch stack to determine whether small pilot flames could heal the voids
 11

observed in unpiloted flames. The hydrogen flows were 0.66 cfs and 0.25 cfs
(1.25 X 10-3 lb/sec)j the methane flow for two pilot flames at the base of the
main flame was 0.0012 crs from ~-inch-od tubing. The greater hydrogen flow is
below the flamedip limit, and both flows are above the f ameback limit. In
motion pictures taken with and without the pilot flames, gaps appeared fre-
quently in the flame, and the flame varied in shape although the flow war,
constant. This indicates that in this present case pilot flames do not con-
tribute appreciably to the healing of holes in turbulent flames, as they do
with turbulent premixed flames in free air (4, !!).

The cycling of flame shape appears to be of the order reported for

smaller flames of other fuels. Barr (1) observed that butane diffusion flames
fluctuate with a frequency of about 10 to 15 hertz. Maklakov (14) also
reports 10 to 15 hertz and finds that the frequency is independent of the
identity of the gas, and that the frequency decreases somewhat with the
increase in burner diameter. For example, fluctuations of a carbon monoxide
flame change from 15 hertz on a 0.5-cm burner to 12 hertz on a 1.2-cm burner.

HIGH FLOW INSTABILITY LIMIT

The high flow instability limit of a diffusion flame, that is, the flow
rate at which tile flame completely leaves the burner port and ceases to exist,
is referred to as the diffusion-flame blowout limit. It parallels the blowoff
limit of premixp.d flames. No blowout limits have been reported previously for
pure hydrogen diffusion flames in still air. Recently Vranos, Taback, and
Shipman (25) reported limits for small hydrogen jets burning in concentric
high-velocity air streams. They found regions of stable burning over a large
range of hydrogen and air velocities, including sonic flows of hydrogen.
Blowout in wind was reported for a 0.1658-inch-id burner.

Blowout is far more complex than blowoff. In earlier work (3), the crit-
ical boundary velocity gradient concept was used successfully to correlate
bloworf limits of premixed laminar and turbulent propane-air flames with the
blowout of turbulent diffusion flames of propane. With this as a precedent,
the concept has been used here to extrapolate from experiments with very small
laboratory burners to very large diameter-flare stacks. Because of insuffi-
cent flow capacity, the blowout of hydrogen diffusion flames from large diam-
eter stacks could not be measured directly in this laboratory. Thus very
small diameter burners had to be used, and even under these conditions blowout
of a neat hydrogen flame was attained only on the smallest orifice used
(table 4). Most of the measurements were for blowout limits of neat mixtures
,
of hydrogen with nitrogen; these data were extrapolated to obtain a blowout
• t limit for hydrogen diffusion flames •

Theory for correlating blowoff limits of premixed flames, using the con-
cept of the critical boundary velocity gradient is reviewed in a previous
report (8). Blowoff supposedly occurs when the local stream velocity every-
where over t'. ? stream cross section exceeds the local burning velocity. Gen-
erally, a flawe stabilizes at the stream boundary where the local flow velocity
and burning velocity are equal. Elsewhere the flame surface assumes an angle
 12

to the local flow direction such that the component of the flow velocity nor-
mal to the flame surface equals the burning velocity. This does not describe
diffusional flame burning because a diffusion flame, particularly a turbulent
one, does not consist of a sharply defined thin boundary separating the
unburned mixture of fuel and air fro~ its burned products. It may be that
this situation is roughly approximated at the base of the diffusion flame over
the burner port and accordingly its blowoff and blowout characteristics can be
correlated by the critical boundary velocity gradient concept.

TABLE 4. - Blowout and lift-off limits of diffusion flames

of hydrogen-nitrogen mixtures

Burner Blowout Lift-off Boundary

diameter, velocity, velocity, Reynolda velocit gradient
inch ft/sec ft/sec No. Sec-l. Average for
blowout. sec- 1
COMPOSITION, PERCENT BY VOLUME: 62.9 R2 , 36.1 N2 , 1.0 O2
0.077 ••••••••••••• 781 - 17,800 7.49 x let 6.09 x let
.179 ...•......... 815 - 42,700 6.36 X let
.233 ••••••••••••• 743 - 51,000 5.13 X let
.306 ••.•••••••••• - 778 69,900 5.15 X let
.306 ............• 797 - 71.600 5.38 X 1(13
COMPOSITION, PERCENT BY VOLUME: 70.2 R2 , 28.7 N2 , 1.1 O2
0.077 ••••••••••••• 853 - 17,200 7.98 X let 8.57 X ICF
.179 ••••••••••••• 925 - 42,900 7.24 X HI
.233 ••••••••••••• 1080 - 65,400 8.97 X loB
.233 ••••••••••••• - 872 53,000 6.23 X loB
-t
.306 ••••••••••••• -
1 1200
957 76,200 6.77 X loB
.306 ............. - 95.700
.. 10.1 X loB
COMPOSITION • PERCENT BY VOLUME . . .
76 8 Rl2 21 7 N~2 , 1 5 0'::I
0.077 ••••••••••••• 820 - 14,400 6.67 X leF 10.8 X IfF
.179 .....•....... - 772 31,300 4.81 X 1(13
.179 •••••••••••.• 1480 - 60.000 1.50 X 107
COMPOSITION, PERCENT BY VOLUME: 87.8 Ra , 10.9 N2 , 1.3 O2
0.077 • • ••••••••••• - 1200 14,900 9.99 X leF 43.5 X 1(13
.077 ••••••••••••• 2760 - 34,300 4.31 X 107
.179 ..•.....••... - 1010 29,100 5.95 X lOS
.179 .•......•.... 13170 - 91,200 4.39 X 107
COMPOSITION, PERCENT BY VOLUME: 100 R2
0.037 .•••••••••••• - 3880 11,200 5.50 X 10"
.037 ••••••••••••• 6730 - 19,400 1.43 X 108
.043 •............ - 3540 12,000 4.40 X 107
,- .043 .............
~.
18920 -
Maximum velocity, blowout not attained.
30,200 2.25 X 1(f

For laminar flow, the gradient 88 is as follows:

88 =8 U/D, (5)

where U is the linear velocity in ft/sec, and D is the diameter in feet. For
turbulent flow, it is as follows:
 .',

13

& = 0.0 39S -U Re .76/

, D, ( 6)

where Re is the Reynolds number.

109r-----~----~----~------~--__. One direct measurement
KEY was made of the blowout
Stack diameter, Inches velocity of hydrogen
() 99.0 (table 4). It was achieved
0 52.0 on the 0.037-inch burner and
!J 31.0 indicates that the critical
X .179 boundary velocity gradient

10 8 •
0
.233
.306
for blowout of a hydrogen
diffusion flame is about 108
reciprocal seconds. Con-
•• Average of 1.0, 2.0, 4.0
.0366 firmation of this measure-
I
en ~ .077 ment by means of hydrogen-
"0 nitrogen mixtures involved
c
0 burners with inside diam-
0
~ eters up to 0.306 inch.
t-= Blowout region Table 4 summarizes all data
~107 obtained with hydrogen-
0 nitrogen mixtures that con-
<l tain over 60 percent
a::
C) hydrogen.
~
..... Stable flame In figure S, extrapola-
U
region tion of the least-squares
9 lines fitting the critical
~106 boundary velocity gradients
~
of table 4 indicates that
a:: the gradient for blowout of
<l a pure hydrogen flame is
0
Z about 10~ sec- 1 • The least-
::>
0 o Aerojet squares line intercepts by
m extrapolation a blowout
point computed on the basis
10 5 of an observation by the
Aerojet-General Corp.,
~Plumbrook
Sacramento, Calif., and
nearly intercepts another
• point observed by the NASA-
Lewis Research Center,
Plumbrook Station: Cleveland,
Ohio. The line exceeds as
102~____~____~~__~______L -_ _~ it should the stable flame
o 20 40 60 80 100 point computed on the basis
HYDROGEN. volume.. ::,ercent of an observation by LASL.
The data leading to these
FIGURE 5. - Blowout limits and Stable Flame Point of three points based on field
Diffusion Flames in Air of Mixtures of experience are in table S.
Hydrogen Plus Inert Gases. Finally, figure 5 shows a
 14

limit obtained with hydrogen-nitrogen mixtures containing about 15 percent

hydrogen. Data for this point will be presented later in discussions of dilu-
tion limits for hydrogen. This lowest point lies below an extrapolation of
the least-squares line, which is to be expected since the blowout curve for
the hydrogen-nitrogen system should intercept the abscissa at some hydrogen
concentration above 4 percent.

TABLE 5. - Blowout limits and stable flame data for diffusion

flames on flare stacks in air of mixtures
of hydrogen plus inert gases

Stack diametel', Hydrogen, Flow.l ft~ /sec Reynolds Blowout

inches percent Hydrogen Nitrogen Steam No. gradient,
sec- 1
1 99 ...............
a 52 ......•......•
19.4
25.6
2,370
2,330
-
2,150
9,830 2.43 X Hf
4,780 1.60 X loB
6.7 X 10'*
2.6 X 106
331 ......••....•.
1.
100 58.1 400 - - 2.43 X 107
Observation by NASA-Lewis Research Center, Plumbrook Station.
3.3 X 107

20bservation by Aerojet-General Corporation, Sacramento.

3Stable flame. Observation by Los Alamos Scientific Laboratory (LASL).

The blowout limit derived in this manner for a hydrogen diffusion flame
should be considered as an order of magnitude evaluation. The laboratory
burners are much smaller than the flare stacks used in the field and one can-
not as yet scale with certainty. That the field experiences are consistent
with laboratory data is reassuring but not necessarily corroborative. Other
reservations about this extrapolation stem from the results reported in a
previous investigation (2).

Examination of figure 5 shows that mixtures which can blowout from flare
stacks because t~e velocity gradient is too much for the particular stack
diameter and the particular hydrogen inert gas composition could conceivably
" be stablized at a lower flow rate or on a larger diameter stack. In other
words, mixtures that do not burn under particular conditions could perhaps
burn under other conditions. Thus, it is a matter of concern whether a
hydrogen-inert gas mixture that is not capable of stabilizing on a particular
flare stack could become hazardous as the flow drifts or is blown else'o1here.

The study of blowout limits also led to a limdted study of flamelift.

This is the condition where the flame lifts off the burner and is stabilized
about I to 2 inches above it (fig. 6). Liftoff limits are included in table 4,
but no correlation is offered. Lifted flames could be blown out by crosswinds
with velocities about 60 to 80 percent of the wind velocities required for
flames that were fully seated on the burner. As shown in table 6, only neg-
ligible quantities of hydrogen were detected below the flame base, except when
the probe was placed into the stream flowing towards the flame base (table 6,
distance from burner axis <0.25 inch).
 15

Seolt, in
20

16

12

FIGURE 6. - Lift-Off of a Hydrogen Diffusion Flame From a D.77-lnch-Diametor Burner.

TABLE 6. - Composition of gases near lifted diffusion flames of

hydrogen-nitrogen mixtures. Burner diameter =
0.179 inch. Flame base = about 0.5 inch diameter.

Height Distance from Velocity, Observed composition of

above burner, burner axis, ft/sec _products) percent
inches inch Hydrogen I Nitrogen 1 O~gen
. . .
COMPOSITION PERCENT BY VOLUME' 76 8 HYDROGEN 21 7 NITROGEN 1 • 5 OXYGEN
) ) )

1.0 .................
1.0 .................
0.19
.38
814
814 I 9.7
.06
74.7
79.3
15.6
20.6
1. 5 ...•......•.....• .50 814 I Trace 79.4 20.5
COMPOSITION) PERCENT BY VOLUME: 87.8 HYDROGEN 10.9 NITROGEN) 1. 3 OXYGEN
1.0 ••••••••••••••••• I
1. O. . . . . . • . . . . . . . • . .
0.19
. 38
2,400
1.940
12.0
Trace
I 70.2
79.5
17.8
20.5

DILUTION LIMITS OR BLOWOUT LIMITS OF HIGHLY DILUTED DIFFUSION FLAMES

OF HYDROGEN-INERT GAS MIXTURES

For various reasons, steady flows of inert gas (nitrogen, helium, or

steam) may be maintained during flare stack operation while h}drogen flows are
varied. This is particularly likely to occur during startup or shutdown of a
run. Conceivably, hydrogen will not be burned when the proportion of hydrogen
to inert gas is too low for diffusion flame burning, and unburned hydrogen will
be discharged from the flare stack. The problem is not one of traditional
flammability limits of premixed static mixtures (1); the blowout limit of a
diffusion flame if. involved. Since blowout is brought about by heavy dilution
with inerts, this limit is termed a dilution limit to differentiate it from
blowout caused by high flow rate of fuel. Theory for correlating blowoff
limits of premixed flames is discussed in the section on the high flow
instability limit and is used here to correlate dilution ~ . imits.
16

Blowout gradients for hydrogen-nitrogen diffusion flames on I-inch,

2-inch, and 4-inch-id burners are presented in table 7; the average values
constitute the coordinates of the lowest point plotted in figure 5. These
are averages of a range of fuel composition ranging from 10.8 percent hydro-
gen to 17.9 percent, and blowout gradients ranging from about 10 to 1700,
variations which are small in 'he context of figure 5. In comparison, the
first two points in table 5 were obtained with full-scale flare stacks flaring
during startup when the hydrogen flow was increasing; the actual limit may be
a little leaner in hydrogen because of time lag between flow readout and
observation of ignition. The difference between the large-scale and labora-
tory runs is not attributable to the different inert gases, because steam and
nitrogen have similar effects on flame stability. The difference in mixture
ratios is due to the difference of the volumetric flows. (Note Reynolds num-
bers.) These data indicate that mixtures that blowout from flare stacks
because the velocity gradient or the dilution with inerts is too much for the
particular stack diameter could conceivably be stabilized at a lower flow rate
or larger diameter. In other words, a hydrogen-inert gas mixture that is not
capable of burning on a particular flare stack could become hazardous as the
mixture drifts, or is blown elsewhere. In contrast to the data in tables 5
and 7, the minimum percent hydrogen in nitrogen among the flammability limits
for the hydrogen-nitrogen-oxygen premixed static system is 4.8 (~). This per-
centage is not to be confused with the lean flammability limit of hydrogen in
air. Incidentally, flammability limit data ar p available in references such
as the one authored by Coward and Jones (1).

TABLE 7. - Dilution limits of diffusion flames in air

of mixtures of hydrogen and nitrogen

Hydrogen, Nitrogen, Hydrogen, Reynolds Blowout

percent ft 3 /sec (X 103 ) number gradient,
~
STACK DIAMETER = 1 INCH
13.6 5.72 0.88 571 116
13.6 9.98 1.58 958 196
15.7 16.5 3.09 1,630 343
15.4 21.4 3.91 2,125 1,022
13.5 29.1 4.56 2,825 1,130
12.7 36.7 5.35 3.530 1.670
STACK DIAMETER = 2 INCHES
12.5 4.87 0.70 244 12
13.4 10.3 1.59 513 26
16.9 15.9 3.24 814 43
STACK DIAMETER = 4 INCHES
10.8 253 29.8 5,930 260
11.5 228 29.2 5,410 215
17.9 57.7 12.2 1,456 19
12.0 119 16.3 2,232 60
15.1 192 34.1 4,700 173
16.7 192 38.3 4,780 179

J.
114.1
Average.
- - - 1365
 17

BURNING RATES

The earlier discussion of fluctuations in flame heights is relevant to

another consideration; namely, whether the reach or len~th of the flame on the
flare stack is sufficient to impinge on the test stand or other facilities.
Two theoretical approaches have been used to predict heights or lengths of
flare-stack flames. The question is whether thef e theoretical treatments can
be effeclive!y applied to 1arge diffusi0n flames. Hawthorne and others (12)
studied flames with Reynol.ds numl °rs up to about 10,000 on burners up to
~ inch in diameter. Wohl and ~thers (26) observed th~t the ratio of flame
height to burner diameter became constant f~r turbulent flames as the Reynolds
number increased; the maximum tube diameter was 0..'. inch, the maximum Reynolds
number was about 32,000. Observations of di.~fus ion flames of high Reynolds
numbers on large-diameter stacks show that fl~me length varies considerably
with flow rate. Putnam's (11) study provides a semiempirical relation between
flame length and flow rate. Thompson and Bonc:are (22) found it necessary to
modify Putnam's empirical constant by a factor of almost 3 for still air in
orde~ to correct for an over-predictton of flame heights.

The approach made in this paper is entirely empirical. Burning rates

have been roughly estimated from photographs of flames on burners ranging from
4 to 31 inch~s in diameter. The burning rate is defined as the volumetric
flow rate of hydrogen divided by the surface area of the flame. Becauoe the
latter is extremely difficult to estimate, only rough determinations can be
made. The results are presented in table 8. Of particular interest is the
difference between apparent burning rp.tes of hydrogen diffusion flames at low
and high flow rates. The burning rates from the laboratory determinations
(about 0.03 and 0.1 ft/sec) are about a'n order of magnitude less than those of
the full-scale flames (about 1 ft/sec), ~hich burned about 1,000 to 10,000 cfs
of hydrogen on stacks about 30 inches in diameter. Turbulence levels were
probably very high. Stack diameters for the two sets of data are within a
factor of 3, and the flows of hydrogen are within factors of three to four
orders of magnitude.

Burning rates can be used to predict flame heights if one assumes that
large-scale diffusion flames have a simple geometric form, such as a frustum
of an inverted cone with a half angle S. Flame height, h, is then given by,

002 tan S + TTDh - V/Su (l-tan2 S)1/2 • 0, (7)

where Su is the burning velocity in ft/sec, V is the flow rate in ft 3 /sec, and
D is the stack diameter in feet. Table 9 shows how flame height depends on S;
a burni~g rate of 1.0 ft/sec was assumed. A half angle of about 3° yields the
best approximation. If a burning rate of 0.1 ft/sec is assumed, no reasonable
value of e (up to 16°) gives as good a fit of the field data. The use of
laboratory-scale data leads to overestimates of flame height.
 ....
OD

TABLE 8. - Burning rates of hydrogen diffusion flames

Stack Flow rate Flame Flame Burnin rate

diameter, Lb/sec Ft~ /sec height, ft surface area, Ft/sec (Lb/sec, ftC!)
inches ft2 X 10-3 X 1~

4 0.16 X 10-3 to 1.6 X 10-3 0.031 to 0.31 1.8 to 4.4 - <0.03 -

12 3.3 X 10-3 .66 3.2 0.0051 .13 0.65

12 4.6 X 10-3 .916 6.2 .013 .07 .36

30 6.6 1,180 62 to 63 .715 to 1.22 1.0 to 1.7 5.4 to 9.2

31 70 12,500 275 to 330 8.27 to 27.1 .5 to 1.5 2.6 to 8.5

TABLE 9. - Predicted heights of hydrogen diffusion flames

versus heights observed in the field

Hydrogen f1ow ••••••••••••••••••• ft 3 /sec •• 1,180 12,500

Stack dia_lllleter .....••.•.•••..•.•...•. in .. 30 31
Observed height •••••••••••••••••••••• ft •• 62-63 275-330
Predicted height, ft:
0° flame spreadl ••••••••••.••.•••••••• 150 1,500
1
3° flame spread •••••••••••••••••••••• 64 254
16° flalllle spread1 •••••••••••••••••••••• 31 108

J. Half angle.
 .' .

19

These burning rates are for diffusion flames, not for stationary or prop-
agating flames of premixed hydrogen and air. A case in point illustrating
propagation rates of large premixed hydrogen-~ir flames is reported by Reider
and others (19). Hydrogen flow rates were about 120 lb/sec, and the authors
estimate that a premixed hydrogen-air flame propagated downward at approxi-
mately 100 ft/sec. In normal combustion, the expansion ratio due to comtus-
tion provides about a sevenfold linear multiplication of the 9 ft/sec burning
velocity. Reider's figure ap,ears reasonable.

Tm1PERATURES OF HYDROGEN DIFFUSION FLAMES

Temperature profiles at various heights above a 4-inch-diameter burner

were recorded by moving an iridium versus iridium-40 percent rhodium thermo-
couple across a diffusion flame. burning 3.5 cfm of hydrogen. The thermocouple
was made of 0.0033-inch-diametpr wire. The flame was about 2 feet high. Tem-
perature fluctuations at a point of measurement ranged from a maximum of
3,590° F to a few hundred degrees Fahrenheit. Average temperatures were
determined by visual estimate based on the app~arance of recorder record~.
Axial average temperatures obtained in this way are given in table 10. Aver-
age temperatures rose from 1,970° F, 1 inch above the port to 2,590° F
13 inches above it and then declined to 1,105° F 25 inches above the port.
The maximum observed temperatJre of 3,590° F may be compared with an adiabatic
fla~ temperature of 3,812° F for a stoichiometric hydrogen-air flame and a
computed diffusion flame temperature of J,OOO° F based on radiation
measurement.

Figure 7 shows the variation of calculated end measured average ~empera­

tures with distance above the flame port; agreement is generally good. Spatial
and temporal temperature fluctuations make it difficult to estimate a single
effective temperature for a hydrogen diffusion flame. The temperature records
show that fluctuations are more frequent near the port wherE' the average tem-
perature is low, and become less frequent at points where th~ average tempera-
ture is higher. For example, 1 inch above the port the t(~mperature fluctuates
700° F above and below the average of 1,970° F about 50 pt!rcent of the time.
However, 9 inches above the port where the average temperature is 2,530° F,
this temperature is exceeded by 700° F less than 25 ~rcent of the time and
does not go above 3,510° F. Temperature measurements also showed that the
iridium versus iridium-40 percent rhodium thermocouple employed does not have
a catalytic effect on the combustion. If a single temperature representative
of a hydrogen diffusion flame must be chosen perhaps toat corresponding to
the peak of figure 7 (about 2,600° F) offers the best compromise, provided
that it is recognized that spot instantaneous temperatures of abo,~~ 3,600° F
can occur.

Samples were collected at the temperature measurement points and analyzed

for combustion products. Table 10 sUDlD8rizes the raw water-free analysis of
products together with computed water-containing products and reactanta which
would yield these products on combustion. In making these ~alculations, the
total number of moles of oxyg~n was computed by multiplying the ratio of oxy-
gen to nitrogen in air by the number of moles of nitrogen foulaJ. From this,
the. water formed was taken to be twice the difference between residual oxygen
20

2,600

2,400

2,200

o o
lL.
0
...
LLJ
0:
:l
~ 1,800
0: KEY
LLJ
a.. • Average measured temperature
~
.--LLJ 1,600
o Adiabatic flame temperature

1,400
o

1,200

20 24 28
DISTANCE ABOVE PORT, in
FIGURE 7. - Temperatures on the Axis of a Hydrogen Diffusion Flame.
Burner diameter =:: 4 inches, flow rate:: 3.5 cfm.

and computed oxygen. The water formed, with the hydrogen observed, gave a
value of reactant hydrogen, and the nitrogen plus total oxygen equaled the
reactant air. Finally, adiabatic reaction temperatures were computed based
on the reactant and product compositions.
 TABLE 10. - Temperatures and analyses of gases on axis of a diffusion flame burning
3.5 cfm of hydrogen on a 4-inch-id burner

Distance Averag p Observed composition Computed composition, Adiabatic

above temperature, of products, mole fraction reaction
burner mout.h, o F mole fraction Products Reactants temperature,
inches H2 N2 O2 H2 N2 O2 H2 O H2 N2 O2 o F

1 1,970 0.404 0.530 0.061 0.349 0.459 0.053 0.139 0.457 0.429 0.114 1,960

3 2,200 .476 .520 .048 .309 .425 .039 .147 .499 .396 .105 2,065

5 2,280 .404 .530 .044 .344 .452 .038 .166 .475 .416 .109 2,280

9 2,530 .260 .688 .062 .208 .549 .049 .194 .366 .501 .133 2,580
1
13 2,590 .066 .814 .125 .056 .685 .105 .154 .195 .636 .169 2,100

17 2,410 .044 .841 .115 .036 .690 .094 .180 .197 .635 .168 2,390

21 1,730 .008 .841 .151 .007 .734 .132 .127 .126 .691 .183 i 1,780

25 1,105 .0006 .82.9 .170 .0005 .754 .155 .091 .088 .721 .191 1,335

....
N
22

Temperatures did not decrease sharply near the axis, as was to be

expected if the fluid there were mainly unburned hydrogen. Instead, tempera-
tures were somewhat higher close to the axis than elsewhere in a given cross-
sectional plane of the flame. The data in table 10 offer a basis for
explaining the decreasing radial temperature gradient in the upper reaches of
the flame, which was observed from the axis to the edge of the flame. The
hydrogen content of reconstructed reactant mixtures along the axis is low and
within the range of flammability (less than 50 pe~cent), showing that air,
flame products, and fuel mix rapidly and that combustion may occur in bulk
rather than at an interface between fuel and air. Moreover, cold hydrogen
may be expected to diffuse outward along concentration gradients; hot water
vapor and hot nitrogen will diffuse inward toward the axis. This mass
exchange within the flame results in an overall transport of heat towards the
axis. At more than 9 inches above the burner mouth, the reactants at the axis
are fuel-lean; they are certainly leaner off the axis than at the axis.
Therefore, in the upper reaches of the flame, increasing radial dilution of
the mixture with distance from the axis causes a corresponding decrease of the
radial temperature gradient.

FLAME CHARACTERISTICS IN WINDS

Stripping of Hydrogen From Its Diffusion Flame by Winds

The possibility was examined that winds may disrupt a hydrogen diffusion
flame and transport unburned hydrogen elsewhere to form a flammable mixture.
Hydrogen was burned on an upright 4-inch-id vertical stack; air at velocities
up to about 40 ft per sec was directed against the hydrogen diffusion flame
from an 8-inch-id horizontal stack. The experiments were conducted indoors to
avoid interference from atmospheric winds. The average position and shape of
the flames were recorded by I-minute exposure photographs, and sampling probes
were installed at selected distances from the flame envelope on the basis of
the flame sha~e determined in this manner.

In a first series of experiments, four single-orifice probes were posi-

tioned around the flame envelope, 1 inch away from it. Samples were withdrawn
within 1 minute and analyzed by gas chromatography with an accuracy of 0.03
percent hydrogen; no hydrogen in excess of this concentration was found in any
of the samples. Table 11 gives the probe positions used to sample hydrogen
flames exposed to various crosswinds. No significant quantities of free
hydrogen were observed.

In a second series of experiments, multiorifice probes sampled from

points spaced 0.5 inch apart around the flame (fig. 8). These integrating
probes were placed at distances ranging from 6 inches outside the flame enve-
lope to 2 inches inside of it. Hydrogen flow from the 4-inch stack was 1 cfs
and the maximum crosswind velocity was 39 mph. Table 12 indicates roughly the
maximum amounts of hydrog!'m found; accurate values could not be obtained
because the detected concentrations were below the accuracy of the gas chro-
matograph employed. Only traces of hydrogen were observed and these just
about vanished at 4 to 6 inches from the photographed envelope. These samples
also show that the hazardous stripping of hydrogen from its diffusion flame by
usual winds is unlikely.
 23

TABLE 11. - Probe positions When sampling around hydrogen diffusion flames
on a vertical 4-inch-id stackl

Height above stack Distance downward Height above stack Distance downward
port, inches from stack axis, port, inches from s tack axis,
inches inches
Hydrogen Flow: 0.082 cfs = 0.94 ft/sec; Hydrogen Flow: 0.33 cfs = 3.7 ft/sec;
Average Air Velocity: 25 ft/sec = 18 mph Average Air Velocity: 39 ft/sec = 27 mph
2 11 25 5
a2 3 12 10
6 2 6 14
4 22 4 26
Hydrogen Flow: 1.0 cfs = 11. 4 ft/sec; 2 0
Average Air Velocity: 39 ft/sec = 27 mph 11 0
13 4 6 5
28 17 2 14
14 17
2 37
l .
No hydrogen detected within analyt~cal accuracy of 0.03 percent hydrogen .
2 Positions below port.

~t--- Airflow 8- in
id

\
To sample To sample
bottle bottle

To sample
bottle ___ H2 stack,
4-inid

FIGURE 8 •• Sampling Equipment for Hydrogen Stripping by Crosswinds.

24

TABLE 12. - Heights of hydrogen peaks on gas chromatographic charts.

(Analyses of samples taken with integrating probes.)

Probe pos ition, At upstream end Above flame At downstream Below flame
distance from of flame envelope, envelope, end of flame envelope,
flame envelope, above burner, inch envelope, inch
inches inch inch
6 Zero Zero 10.06 Zero
4 Zero 0.06 .OS Zero
2 0.07 .11 .14 0.09
1 Zero .06 .07 .OS
22 Zero .OS .11 .11
J. .
" (accur~cy of analys1s) = peak height of O.S inch •
0.0370
2 Probe is inside flame.

Blowback of Hydrogen Diffusion Flames by Winds

The possibility of flame being blown back into a horizontal flare stack
by wind and thus damaging the stack was also investigated. An array of ther-
mocouples was placed in a horizontal 4-inch-id stack through which hydrogen
flowed. This stack was faced by a horizontal S-inch-id stack, from which an
opposing stream of air flowed. The thermocouples were used to measure temper-
atures at a number of points under varying conditions of opposing air and
hydrogen flows. Temperatures were recorded by a fast-response oscillograph.
Average temperatures recorded ranged up to 1,900° F, with transient maximum
temperatures about 150° F higher. Because the melting point of stainless
steel is about 2,700° F there is no danger of a water-cooled stainless steel
duct melting because of a hydrogen diffusion flame being blown into it by
~· inds. Figures 9-11 show the average temperatures as a function of distance
into the stack from its port. Except for the lowest hydrogen flow (fig. 9),
the temperatures in the stack decrease monotonically from ~he port towards the
base of the stack. For the lowest hydrogen flow, a local maximum temperature
occurred 3 inches down the stack. For the highest air flow and the same
hydrogen flow, the temperature at the 3-inch level was about the same as at
the port (fig. 9). Although these temperature profiles do not clearly locate
the flame position, they are significant in establishing the temperature
levels that may be expected inside a hydrogen flare stack facing into an
opposing wind. The possibility that the temperatures observed were affected
by nonuniform wall temperatures was discounted by showing that the outside
temperature of the first 6 inches of stack wall was about 1,600° F all around
the tube. The temperature profiles in figures 9-11 are therefore longitudinal.

Samples taken at points where temperatures were measured were analyzed

for combustion products (tables 13-14). The raw water-free analysis of the
products is given, together with the computed water-containing products and
the reactants that would yield these products on combustion. The rea~tant
compositions show that when the hydrogen velocity was 12 ft/sec against an
opposed wind of 45 ft/3ec, the compositions were flammable to a depth of 4
inches but not at depths of 5 and 6 inches (table 13 and fig. 12). On the
other hand, when a hydrogen velocity of 3S ft/sec was opposed by a wind veloc-
ity of 42 ft/sec, flammable compositions extended only as far as 2 inches,
 t I I •

25

2,000----------~--------~----------~--------~

Air velocity. 42 ft/sec-...,

1,500

o
.
~ Air velocity, 30 ft/sec

~a:::
~ 1,000
~ Air velocity, 10 ft Isec
~
w
(!)
<t:
a:::
w
~
500

Hydrogen velocity, 12 ft Isec

° 2
DISTANCE FROM PORT, inches
3 4

FIGURE 9. - Average Temperature Observed Inside a 4-lnch-ID Stack Burning at a 12 ft/sec

Hydrogen Velocity and Facing Into Air Blast.

with compositions at the 3- and 4-inch levels being in the nonflammable range
(table 14 and fig. 12). Visual observ~tion indicated that the flame was blown
26

2,000-----------~-------r----........,

Air velocity, 42 ft /sec

1,500
lJ...
o

UJ
..
0::
::)
~ Air velocity, 30 ft/sec
0::
UJ
~ 1,000
UJ
I-
w
(!)
<t
a::
w
>
<t

500
Ai r velocity, 10 ft /sec/

Hydrogen velocity, 22 ft/sec

o 1 2 3 4
01 STANCE FROM PORT, inches
FIGURE 10 .• Average Temperature Observed Ir.side a 4·I"nch·ID Stack Burning at u 22 it/sec
Hydrogen Velocity and Facing Into Air Blast.

into the stack about 4 to 6 inches, f~r a hydrogen flow of 12 ft/sec and about
3 inches for a flow of 38 ft/sec.
'"
 TABLE 13. - Temperatures and analyses of gases inside of a horizontal 4-inch-id
diffusion flame burner. facing into an opposing wind. Hydrogen
velocity. 12 ft/sec; axial air velocity, 45 ft/sec.

Distance Observed1 Observed Computed composition of Computed

from burner average composition of products, mole fraction Reactants, mole adiabatic
mouth, temperature, products. mole reaction
inches o F H2 N2 O2 H2 N2 O2 H2 O H2 N2 O2 temperature
o F
2 1,485 0.198 0.700 0.054 0.163 0.576 0.045 0.216 0.379 0.576 0.153 2,800
3 1,945 .598 .382 .005 .508 .325 .004 .163 .671 .325 .086 2,270
4 1,330 .594 .362 .014 .524 .319 .012 .145 .669 .219 .085 2,040
4 1,330 .632 .318 .029 .580 .292 .027 .101 .681 .292 .078 1,500
5 - .804 .180 .010 .752 .168 .009 .071 .823 .168 .045 1,090
6
Separate run.
- .864 .110 .008 .843 .108 .008 .041 .884 .108 .029 690

TABLE 14. - Temperatures and analyses of gases inside of a horizontal 4-inch-id

diffusion flame burner. facing inte Qn opposing wind. Hydrogen
velocity. 38 ft/sec; axial air velocity. 42 ft/sec.

Distance Obserred Observed Computed composition of Computed

from burner average composition of products, mole fraction Reactants, mole adiabatic
mouth, temperature, .products ~ mole reaction
inche s o F H2 N2 O2 H2 N2 O2 H2 O H2 N2 O2 temperature
o F
1 1, 440 0.180 0.652 0.090 0.166 0.600 0.083 0.151 0.344 0.652 0.172 2,080
2 815 .5~: .356 .017 .525 .321 .015 .139 .664 .321 .085 1,970
3 - .820 .148 .012 .793 .143 .012 .052 .845 .143 .038 836
3
4
-35 .802
.878
.182
.102
.011
.013
.751
.860
.170
.100
.010
.013
.069
.027
.820
.888
.170
.100
.045
.027
1,080
330

N
-...J
28

1,500-------r-------'T-----~------,

Hydrogen velocity, 38 ft /sec

Air velocity, 42 ft /sec

u..
o
...
~ 1,000
~
0:
W
0...
r5 Air velocity, 30 ft/sec
t-
w
~
0::
w 500
~

Air velocity, 10 ft /sec

o 2 :3 4
DISTANCE FROM PORT, Inches
FIGURE 11 .• Average Temperature Observed Inside a 4·lnch·ID Stack Burning at a 38 ft/sec
Hydrogen Velocity and Facing Into Air Blast.
The reactant and product compositions were used to compu ~ e adiabatic
reaction temperatures. Calculated adiabatic flame temperatures were consid-
erably higher than the experimental temperatures. This is to be expected
because the calculations assume that there is no heat loss to the walls. As
such, they may be considered to set an upper limit to the temperatures pro-
duced by hydrogen diffusion flames blown into a duct. Only at the 2-inch sam-
pling point with low flows of hydrogen was the compvted temperature high
enough to damage an uncooled stainless steel flare stack. Apparently blowback
of flames under conditions comparable to these will not endanger water-cooled
flare stacks.
 29

100~------------~-----'------~------r-----~

...c / 0
C»
/
~ 0
/
C» /
~
Q.
I
C» 80 Upper flammability limit /
E
::J
'0
/
>
. /
z
0
~
<t
~ /
/
cf • •
zIJJ 60 /
U
z /
0
u /
z
IJJ
/
(!) I KEY
~
c
I H2 velocity, 12 ft/sec
>- I •
I: 40 I Air velocity, 45 ft/sec
IJJ
> I ~ velocity, 38 ft/sec
~
u
10 0
1 A ir velocity, 42 ft/sec
'. IJJ
IJ..
IJ..
IJJ
I
I
20
2 3 4 5 6
01 STANCE FROM PORT, inches
FIGURE 12 •• Effective Hydrogen Concentration in 4·lnch·ID Stack With Diffusion Flame
Bum ing Against Opposed Wind •
.
.
Experiments were performed to determine the position of the stagnation
point of the wind blowing into the cavity formed by ~he stack, wh.ch was
effectively closed at the ,o pposite end. Static and total pressure heads were
measured by Pitot tube at various points inside and outside of the 4-inch-id
stack closed at the far end; there was no hydrogen flow, and wind wa ~ blowing
against t he stack with an average velocity of 58 ft/sec (table 15). Readi~g.
are tim.! averages taken over a period of about 1 minute. Slow-response slope
gages we l ~ used to measure the pressure heads. The readings provide no
30

information about instantaneous velocities or directional changes of the flow.

Only the rough coincidence of flame position and the location of the stagna-
tion point of the wind blown into the stack are considered significant.

TABLE 15. - Total and static pressures and corresponding air velocities
produced by a wind from an 8-inch-id duct blowing against
a coaxial 4-inch-diameter horizontal stack

Distance Radial distance Total Static Calculated

inside from top edge of pressure, pressure, air
4-inch stack, 4-inch stack, inches of water inches of water velocity,
in. in. ft/sec
.L~ 0.25 1.2 .L 0.3 77
1.25 1.4 1 .3 84
2.25 1.5 1.3 86
3.25 1.6 1.3 89
13 .25 1.1 1 .1 71
1.25 1.5 0 77
2.25 1.9 .1 86
3.25 1.9 0 89

0 .25 1.2 .8 41
1.25 1.7 1.6 :!1
2.25 2.0 1.6 41
3.25 1.8 1.6 29

2 .25 2.5 2.4 21

1.25 2.3 2.2 21
2.25 2.6 2.2 41
3.25 2.6 2.1 46

4 .25 2.6 2.6 0

1.25 2.5 2.5 0
2.25 2.6 2.4 29
3.25 2. 6 2.6 0

6 .25 2.6 2.8 1 29

1.25 2.7 2.8 121
2.25 2.6 2.8 1 29
3.25 2.7 2.8 121

8 .25 2.7 2.8 121

1.25 2.7 2.5 29
2.25 2.5 2.8 135
3.25 2.7 2.8 121
lDistance outside 4-inch stack. Distance between 8-inch duct and 4-inch
stack • 12 inches. Average air velocity from 8-inch duct • 39 mph •
58 ft/sec.

"'-' -' - - -
 31

Thoug'n no hydrogen flo\\l was Ilsed in these experiments, the location of

the stagnation point is probably the same as with a low hydrogen flow; tre
velocity pressure heads due to low flows of hydrogen are much less than the
velocity pressure h~ad of the winrl in these experiments. The per.k velocities
reported In table 15 agree roughly with velocities expected from flowmeter
" · readings; the latter correspond to an average velocity of 58 ft/sec from the
8-inch ciuct. Assuming that the flow profile was about that for turbulent pipe
flow (12), the velocity in the axial core of the stream would be about
70 ft/sec. The agreement was good between velocities based on the two types
of measurements, suggested that the stagnation distance was located correctly.
Thus the flame tends to position itself at about the stagnation point of the
wind blowing into the stack.

Another conceivable interpretation is that the distance that the flame is

blown back into the stack by wind is determined by the compression of the on-
coming hydrogen and that the maximum flame penetration can be predicted f~om
the perfect gas law for a supposed stagnall:-- column of hydrogen. According to
this concept, the depth of the penetration would depeud on the length of the
stack and not on its dia~ter. However, this concept was shown to be invalid
by experiments witt. 2-inch- and 4-inch-id stacks, 3 feet long. Flame Pi.l!"!,C-
trated deeper into the 4-inch stack than into the 2-inch stack when air veloc·.
ities were the same for both sets of experiments, and the hydrogen flow \laS
40 ft/sec through the 4-inch stack and 23 ft/sec through the 2-inch stack; the
flame penetrations were 3 inches and ~ inch, respectively. Computations based
on velocity heads and friction with the stack walls did not correlate with
actual flame positions.

Bl~wout Limits of Hydrogen-Nitrogen Diffusion Flames in Crosswind~

An earlier ~ection of this paper dealt with blowout limits in still air
of flames of hydrogen and nitrogen mixtures heavily diluted with nitrogen;
ratios of hydrogen to nitrogen at. blowout were determined at various flows of
nitrogen. Since flows w~th high concentrations of inerts are often encoun-
tered in the vperation of flare stacks it was reasonable tn examine the
increase due tc a crosswin.:i in the hydrogen-ine:'t gas ratio required for
stable burni;.! 3. Experiments were performed in which winds of various veloci-
ties hllpingec! latt~rally on a diffusion flame from an upright 4-inch stack
(table 16). Within the experimental uncertainty, the blowout ratios of flames
of these nlixtu=~s burning in winds of up to 50 ft/sec are not significantly
differpnt. from t:l1~se in still air. It was observed, however, that the wi.nds
drove flames into the stack as much as a foot from the port.

These three studies of hydrogen diffusion fla~ chrracteristics show that

winds do not cr.eate an additional hazard in fl3re st uck operation attrib~table
to stripping of hydrogen from the flame, blowing of flame into the stack, or
excessive instability of hydrogen-inert gas flames.
32

TABLE 16. - Effect of crosswind on blowout limits of hydrogen-nitrogen

diffusion flames on a 4-inch-diamet er stack

Average wind l:ydrogen, Stack flow. ft 3 /sec Reynolds Velocity,

velocity, percent Nitrogen Hydrogen No. ft/sec
ft/sec
10 8. 7 0.17 ':' 0.0267 4100 2.29
12.7 18.0 .117 .0258 2900 1.64
15.3 15.3 .147 .0267 3530 1.99
20.5 15.3 .152 .0266 3680 2.04
22.5 14.5 .159 .0272 386 : ~ 2.13
30.8 15.3 .146 .0260 3550 1.97
32.3 15.3 .148 .0267 3590 • 2.00
41.6
45.1
18.0
16.7
.117
.130
.0259
.0262
2900
3190 I 1.64
1. 79
48.5
54.5
.
1 Averages 1nclude
16.7
16.0
.
.132
.142
.
.0260
.0263
data prev10usly reported 1n table 7 .
3230
344u I 1.81
1.92

Smt-tARY

The safe operation of flare stacks raises problems of flame instability

due to (1) low and high flows of hydrogen; (2) dilution of hydrogen with
inerts; (3) heating hazards related to the temperature of the flame and to its
length which in turn is related to its burning rate; and (4) the effect of
winds upon flame shape, flame position, and the conceivable stripping of
unburned hydrogen. Laboratory experiments supplemented with some field expe-
rience prcvide data that can be used to anticipate the existence and degree
of these hazards. For example, considering a 52-inch-diameter flare stack
the minimum flow required to avoid flamedip is predicted to be about 0.18
1bs/sec of hydrogen. The minimum flow at which blowout would occur from an
l8-inch stack is predicted to be about 140 lbs/sec of hydrogen. Temperatures
across large hydrogen diffusion flames ~ay approach about 3,600° F, but a more
typical value is about 2,600° R Other spec if ics such as flame height can be
computed when operational parameters, such as flow, are known.
 33

REFERENCES

1. Barr, John. Diffusion Flames. Paper in 4th Symposium (International) on

Combustion. Williams and Wilkins Co., Baltimore, Md., 1953, pp. 765-
771.

2. Bluhm, W. C. How To Operate a Flare System Safely. Oil and Gas J.,
Aug. 28, 1961, pp. 73-77.

3. Bollinger, L. M., and D. T. Williams. Experiments on Stability of Bunsen

Burner Flames for Turbulent Flow. Nat. Advisory Comm. Aeronaut. Tech.
Note 1234, June 1947, 18 pp.

4. Burgess, D. Structure and Propagation of Turbulent Bunsen Flames.

BUMines Bull. 604, 1962, 42 pp.

5. Coward, H. F., and G. W. jones. Limits of Flammability of Gases and

Vapors. BUMines Bull. 503, 1~52, 155 pp.

6. Gordon, A. S., S. R. Smith, and J. R. McNesby. Study of the Chemistry of

Diffusion Flames. Paper in 7th Symposium (International) on Combustion,
Butterworths, London, 1959, p. 317.

7. Grumer, J. Flashback and Blowoff Limits of Unpiloted Turbulent Flames.

Jet Propulsion, v. 28, November 1958, pp. ,56-758.

8. Grumer, J., M. E. Harris, and V. R. Rowe. Fundamental Flashback, Blowoff

and Yellow-Tip Limits of Fuel-Gas-Air Mixtures. BUMines Rept. of Inv.
5225, 1956, 199 pp.

9. Hajek, J. D., and E. E. l ' ldwig. How To Design Safe Flare Stacks. Petrol.
Chem. Eng., v. 32, June 1960, pp. C3l-C38; July 1960, pp. C44-C51.

10. Hawthorne, W. R., D. S. Weddell, and H. C. Hottel. Mixing and Combustion

in Turbulent Gas Jets. 3d Symposium on Combustion, Flame and Explosion
Phenomena. Williams and Wilkins Co., Baltimore, Md., 1949, pp. 266-288.

11. Karlovitz, B. , D. W. Dennison, Jr., D. H. Knapschaefer, and F. E. Wells.

Studies on Turbulent Flames. Paper in 4th Symposium (International) on
Combustion, Williams and Wilkins Co., Baltimore, Md., 1953, pp. 613-620.

12. Lapin, A. Hydrogen Vent Flare Stack Performance. Ch. in Advances in

Cryogenic Engineering, Vol. 21. Plenum Press, New York, 1967, pp. 188-
206.

13. Lewis, B., and G. VOl, Elbe. Stability and Structure of Burner Flames.
J. Chem. Phys., v. 11, 1943, pp. 75-97.

14. Maklakov, A. I. Oscillation of Diffusion Flames Appearing When There Is

a Laminar Flow of the Ignitable Medium. Zh. Fiz Khim., v. 30, 1956,
p. 708.
34

15. Peterson, P. Explosions in Flare Stacks. Chem. Eng. Prog., v. 63,

No.8, 1967, pp. 67-70.

16. Prandtl, L., and O. G. tietjens. Applied Hydro- and Aeromechanics.

Dover Publications, Inc., New York, 1957, pp. 48-49.

17. Putnam, A. A., and C. F. Speich. A Model Study of the Interaction of

Multiple Turbulent Diffusion Flames. Paper in 9th Symposium (Inter-
national) on Combustion. Academic Press, New York, 1963, pp. 867-877.

18. Reed, R. D. Factors for Safe Design of Operation of Flares and Flare
Systems. Pres. at the Symposium on Explosion Effects and Flare Stacks.
63d Nat. Meet. AIChE, Feb. 18-21, 1968. American Institute of Chemical
Engineers, New York, 1968, pp. 1-15.

19. Reider, R., H. J. Otway, and H. T. Knight. An Unconfined Large-Volume

Hydrogen/Air Explosion. Pyrodynamics, v. 2, 1965, pp. 249-261.

20. Singer, J. M., and J. Grumer. Carbon Formation in Very Rich Hydrocarbon-
Air Flames. BUMines Rept. of Inv. 6007, 1962, 80 pp.

21. Smith, S. R., and A. S. Gordon. Studies of Diffusion Flames. I. The

Methane Diffusion Flame, J. Phys. Chem., v. 60, 1956, p. 759.

22. Thompson, W. R., and C. S. Boncare. Design aud Development of a Test

Facility for the Disposal of Hydrogen at High Flow Rates. Paper Q-6 J
Cryogenic Engineering Conf., June 13-15, 1966, Boulder, Colo., 12 pp.

23. Von Elbe, G., and J. Grumer. Air Entrainment in Gas Burners. Ind. and
Eng. Chem., v. 40, No.6, June 1948, pp. 1123-1129.

24. Von Elbe, G., and M. Mentser. Further Studies on the Structure and
Stability of Burner Flames. J. Chem. Phys., v. 13,1945, pp. 89-100.

25. Vranos, A., E. D. Taback, and C. W. Shipman. An Experimental Study of

the Stability of Hydrogen-Air Diffusion Flames. Combustion and Flame,
v. 12, June 1968, pp. 253-260.

26. Wohl, K., C. Gazley, and N. Kapp. Diffusion Flames. Third Symposium on
Combustion, Flame, and Explosion Phenomena. Williams and Wilkins Co.,
Baltimore, Md., 1949, pp. 288-300.

INT.-BU.OF MINES,PGH.,PA . ISSS4