VENTING OF LOW PRESSURE HYDROGEN

GAS
A Critique of the Literaturey

G. R. Astbury

Health and Safety Laboratory, Buxton, Derbyshire, UK.

Abstract: The use of hydrogen is increasing in industrial processes, and its use is likely to
increase further with its potential use as an energy carrier. The venting of hydrogen is inevitable
at some time for almost all uses, and its propensity to ignite makes it essential that safe venting
regimes are understood. The options for disposal by dilution to below the lower flammable limit,
inerting, dispersion of flammable concentrations and flaring are discussed, along with the poten-
tial for ignition of releases within the flammable range and the subsequent need for flame extin-
guishing. The available literature on the protection of vents from external ignition is critically
examined, to determine the appropriate parameters to allow selection of a disposal method.
A decision-tree is presented to allow a rational appraisal of the most appropriate disposal
method to be selected and precautions to be applied for adequate protection of the vent.
Keywords: hydrogen; venting; explosion; fire; inerting; safety.

INTRODUCTION as the vessel being de-pressurized eventually

reaches atmospheric pressure, there is a ten-
There is an increasing number of processes dency for air to enter the vent pipe, and pro-
being operated where hydrogen is released. duce a flammable mixture within it, thus
Such releases can be of several forms, such presenting the potential for a confined
as a short decaying flow as a hydrogenator explosion to occur. As a consequence, it has
is de-pressurised or vented down; a continu- been common practice to inert gas purge
ous flow as in a reaction releasing hydrogen; the vent line to prevent the ingress of
or a combination such as a steady release of oxygen. Purge regimes and their calculation
hydrogen during a reaction, followed by a are discussed later.
decaying flow as the batch work-off proceeds. Where reactions produce hydrogen as a


Correspondence to: The major hazard that occurs with hydrogen by-product, the ideal solution is to use the
Mr G.R. Astbury, Health and is its ease of ignition coupled with its extreme gas as a fuel. However, as it is invariably
Safety Laboratory, Harpur flammability, resulting in a potential for impractical to collect and transport the gas, it
Hill, Buxton, Derbyshire,
explosion. As hydrogen has a high burning is usual to simply vent it. For reasonably
SK17 9JN, UK.
E-mail: graham.astbury@ velocity, the potential for detonation arises small flows it is practical to simply vent, but
gmail.com especially if the explosion is confined to any it is necessary to assume that ignition is
degree. However, mitigating effects are that inevitable at some time and either design
DOI: 10.1205/psep06054 it readily disperses because of its buoyancy, the vent to cater for an ignition or dilute the
0957–5820/07/
and that its flame has a low emissivity. gas to be non-flammable. Where the flows
$30.00 þ 0.00 are large, it would have to be burnt in a
proper flare stack.
Process Safety and
DISCHARGES
Environmental Protection
Venting
Trans IChemE
Relief Valves
Where hydrogen is deliberately released
Part B, July 2007 The case of the infrequent release from a
infrequently, such as at the end of a hydro-
# Crown Copyright 2007 genation, the quantity released is usually relief valve needs to be considered. Under
small and the release time is short. Under these circumstances, the relief valve is held
these circumstances, the hydrogen may open by the pressure being above the set
simply be vented to atmosphere. However, pressure. As the flow decreases, the valve
will start to close, finally closing fully. Under
†
Crown copyright 2007. This article is published with
normal circumstances, this produces a leak-
the permission of the controller of HMSO and the tight metal-to-metal seal with minimal flow of
Queen’s Printer for Sctoland. any hydrogen. Under some circumstances,

289 Vol 85 (B4) 289–304

290 ASTBURY

a soft-seated valve may be used to ensure bubble-tight shut- disperse with minimal flammable volumes even under pas-
off. Thus under relief conditions, the gas flow velocity will be sive ventilation in enclosed spaces. Other analytical work
high and well above the flame speed, so burn-back will not has been undertaken by Webber et al. (1997) on the dis-
occur. As soon as the relief valve closes, the flow will stop, persion of very buoyant plumes, and they conclude that
and any flame that may be present will start to burn back simple one-dimensional models may be adequate for deter-
down the relief valve tail-pipe. However, whilst this may mining the concentration profiles during dispersion of such
occur, providing the pipe can withstand any detonation plumes, so for high hydrogen concentrations, their work
pressure, the flame will be unable to propagate past the may well be applicable, whereas it may not be at high inert
valve seat because of the very close contact of the disc on gas concentrations. Long (1963) has reviewed some simple
the seat. Therefore no purge would be required for a relief published work on estimation of the extent of hazardous
valve providing the pipe-work can withstand the pressure of areas around vents, and other guidance is available (BSI,
any possible detonation that may occur. A typical maximum 2003; Institute of Energy, 2005), although these are not
detonation pressure would be of the order of 30– 40 bar, so specifically aimed at hydrogen. Once the extent of any flam-
Class 300 pipe-work should be adequate. However, if a mable volume of the hydrogen has been established, then
balanced bellows relief valve were to be used, then the bel- the consequence of its ignition will have to be considered.
lows may be unable to withstand the excess pressure of The quantity of hydrogen in the flammable zone will give an
the detonation on the discharge side, and an inert gas indication of the energy release, bearing in mind the heat of
purge would be required to avoid propagation of flame combustion of hydrogen in air is 120 MJ kg21 (Barnett and
down the relief valve tail-pipe. Hibbard, 1957). The potential for a hydrogen –air cloud to
Where a bursting disc is used as the means of relief then, detonate is known but the criteria required are recognised
since it is not self closing on loss of pressure, it will be as a strong ignition source, a sufficiently large cloud for the
necessary to protect the tail-pipe of the vent with an inert combustion wave to propagate through, and some form of
gas purge as for a normal process vent. turbulence induced by obstacles or other perturbation in the
combustion wave (Cadwallader and Herring, 1999). For
small or slow releases of hydrogen, which is buoyant, the
Dispersion
likelihood of detonation is remote providing that the stack
The standard work on gas dispersion, prior to commercial emitting the hydrogen is not within a confined or congested
computational fluid dynamics (CFD) programmes being avail- area. As hydrogen disperse readily, it is likely that a release
able, suggests a minimum efflux velocity for a flammable gas which ignites will not consume all of the hydrogen, as some
in order to dilute it adequately before it loses momentum and will have been diluted below the lower limit, depending on
descends to the ground as a flammable cloud (Cude, 1974). the time delay between ignition and the release. In one inci-
Clearly hydrogen does not have sufficient density to fall when dent where a release of hydrogen ignited some 12 s after the
mixed with air, and its low density means that the calculated release at 55 kg s21, the flame front travelled at about
required minimum efflux velocity would exceed the velocity of 30 m s21, which was some two orders of magnitude less
sound in hydrogen, which is over three times the velocity in than a detonation but was about one order of magnitude
air. At such a high velocity the flow would be likely to ignite faster than would be expected of a normal deflagration
(see section on ignition below), and the flame would be extre- (Reider et al., 1965). The total quantity of hydrogen con-
mely noisy, but because of the low density, high buoyancy sumed was estimated, from the damage caused, at about
and high diffusivity of the hydrogen, it disperses readily with- 10% of the 900 kg released in total, showing that a majority
out a high efflux velocity. Where a considerable quantity of of the hydrogen had dispersed and diluted prior to ignition.
hydrogen has to be vented, it is necessary to ensure that it Where hydrogen is not deliberately flared, the over-
is adequately dispersed. As the hydrogen is vented, it will pressure produced by the ignition of the flammable zone of
mix with the air in the atmosphere surrounding the vent, the hydrogen– air mixture can be estimated. One suggested
and form a flammable atmosphere. If this is allowed to simple method (Health and Safety Commission, 1979) is
accumulate, there is the potential for an unconfined gas the TNT Equivalence Method, which is detailed in the
explosion to occur. Such explosions have occurred in the I. Chem. E. Monograph (Institution of Chemical Engineers,
past, without direct injury, but a man fell off a scaffolding 1994). Alternatively, more recent commercially available
with fright when such a gas explosion occurred, according computer simulation packages can also be used. If the calcu-
to one source (McBrien, 1988). lated blast effects cause no potential problems to surrounding
However, if the vent is in close proximity to a building, or buildings, then an ignition of the hydrogen can be tolerated.
has a large flow, then it would be necessary to dilute it with Clearly if the adjacent buildings are too close, then the only
inert gas or deliberately flare it. The flow at which flaring options are to relocate the vent further away from buildings,
becomes desirable is open to debate, and each case would or dilute the vent to make the hydrogen non-flammable on
have to be decided on its own merits. mixing with air.
The height of stack required to effectively disperse the
hydrogen is arguable. Cude (1974) recommends calculating
the concentration down-wind of the stack, and adjusting the EXPLOSION AND PREVENTION
height if the concentration is too high for the neighbouring
Ignition
buildings. The use of a positive buoyancy in his equations
will enable the extent of the flammable zone to be deter- Hydrogen has very wide flammable limits of 4–75% v/v,
mined. Alternatively, commercial CFD simulations can be and a low minimum ignition energy of 0.017 mJ (ISO,
used. Work by Swain and Swain (1992, 1996) indicates 2000). The low ignition energy (less than 10% of a typical
that small leaks of hydrogen are so buoyant that they hydrocarbon), makes it so sensitive to ignition that basing

Trans IChemE, Part B, Process Safety and Environmental Protection, 2007, 85(B4): 289– 304
 VENTING OF LOW PRESSURE HYDROGEN GAS 291

safety on the avoidance of ignition sources is almost imposs- The critical voltage to initiate a corona discharge for a 10 m
ible. Consequently, it has to be recognised that where hydro- high stack with an effective edge radius of say 1 mm would be
gen is vented, and can mix with air to form a flammable pffiffiffi pffiffiffiffiffiffiffi
atmosphere, it will inevitably ignite at some time. An ICI Vs ¼ 18 r ¼ 18 0:1 ¼ 5:69 kV (5)
Safety Newsletter (Anon, 1972) discusses the problem, and
points out that stacks which emit flammable gases which This is greater than the potential occurring during fine weather
are not deliberately flared can still ignite, and suitable conditions so therefore a corona would not form, but under
means of prevention of propagation of flame back to the conditions of falling snow (Burrows and Hobbs, 1970), the
plant is required. field strength can rise to 6.6 kV m21. Thus a corona discharge
The ignition of vent stacks is usually associated with may then occur, igniting the hydrogen. Under exceptional local
atmospheric electrical activity, which most people recognize conditions, the field strength can rise to typically 3 MV m21,
as thunderstorms. The Safety Newsletter advises that some and this can initiate a spark discharge (i.e., a lightning strike).
plants should increase the flow of inert gas when there is a Note that the potential difference is proportional to the stack
lightning warning. However, because of the sensitivity of height, so it can be seen that increasing the stack height to
hydrogen to ignition, an electrical storm is not necessary to aid dispersion will increase the potential, and hence the pro-
initiate combustion. A corona discharge from the end of the pensity to ignition. This is borne out in practice as fitting
stack is sufficient to cause ignition. Corona discharges are higher stacks to hydrogen vents resulted in more frequent igni-
present when there is an increased electrical field strength tions, even though hydrogen discharges were less frequent in
in the air. Where an electrode (in this case the stack) has a one case (Sellers, 1977).
potential difference from the surroundings, the potential at In stormy weather, the gradient can approach 20 kV m21,
which a corona discharge will initiate depends on the magni- so the local field strength in the case of the 1 mm radius will
tude of the smallest radius. Cross (1987) gives the threshold exceed 3 MV m21, and a corona discharge is bound to
field that must be exceeded at a discharge point of radius occur. Thus ignition would appear to be inevitable. On frosty
r before a corona is initiated as nights, when the hoarfrost descends from the cold night sky
towards the ground, and in sleet and falling snow, significant
18 charge is carried on the ice particles (Camp, 1976), thus
Ec ¼ (1)
r 1=2 increasing the potential gradient well above the 100 V m21
normally encountered. Thus it would seem likely that any
where Ec is the threshold field, kV cm21; r is the radius, cm. vent emitting hydrogen is certain to ignite at some time. The
Calculation of the potential difference required to produce experience of two adjacent chemical companies at Ellesmere
the threshold field is complicated in all but the simplest of Port (Bradburn and McBrien, 1983) indicates that many igni-
geometries. In the case of a stand-alone stack, the tip of tions appeared to be associated with sleet and falling snow.
the stack can be assumed to be a single discharge point
and the base of the cloud layer in the atmosphere above
can be assumed to be a plane surface. As a first approxi- Avoidance of Ignition
mation, it can be assumed that the potential which must be However, the ignitions are not as frequent as the above
applied between the single discharge point and a plane to would suggest. The possible explanation behind this is that
produce a field Ec at the surface, is independent of the the hydrogen emitted has to mix with air, and that the very
spacing to earth and is given by low ignition energies are only associated with near-
stoichiometric conditions. Hence in a stack venting a large
Vs
Ec ¼ (2) volume of hydrogen, the near-stoichiometric conditions will
r only apply some distance from the end of the stack, and
where Ec is the field strength at which breakdown occurs, hence away from the potential corona discharge point.
kV cm21; Vs is the potential of the electrode, kV; r is the Lower flows are therefore more likely to be both within the
radius of curvature, cm. flammable range and close to the corona point.
Combining equations (22) and (23), the voltage required is A Technical Note (Ministry of Technology, 1968) suggests the
given by fitting of a toroidal anti-corona ring thereby to increase the
effective radius of a stack exit, presumably, increasing
pffiffiffi the potential at which a corona discharge occurs. The note
Vs ¼ 18 r (3)
states ‘. . . a toroidal ring is fabricated to the dimensions of
with Vs in kilovolts and r in centimetres. Since this voltage is the vent stack outlet and welded to it. This ring inhibits the
very localised, it is the effective potential to initiate a corona dis- flow of current (sic) at the stack lip by removing the cause of
charge. If the electric field strength exceeds about 3 MV m21 turbulence characteristic of a sharply defined vent exit . . .’.
(i.e., 30 kilovolts cm21), then total breakdown occurs, and This would appear to suggest that the turbulence induced by
the result is a spark discharge i.e., a lightning strike. Normally a high pressure discharge generates the electrostatic
in fine undisturbed weather (Sellers, 1977) there is a potential charge, but this would be very unlikely for low pressure
gradient of about 100 V m21. A stack 10 m high would be at a releases unless there were entrained particulates in the
potential difference from the electric field of vented gas. However, the Technical Note later states ‘. . .
tests have shown that hydrogen gas at flow rates no greater
V then 1.0 lb s21 (0.454 kg s21) may be safely vented through
Potential ¼ gradient  height ¼ 100  10 m stacks averaging 30 to 40 ft (9 to 12 m) in height without
m
anti-discharge modification . . .’. In practice, the effect of such
¼ 1000 V ¼ 1 kV (4) an ‘anti-discharge ring’ is unlikely to be maintained for anything

Trans IChemE, Part B, Process Safety and Environmental Protection, 2007, 85(B4): 289 –304
292 ASTBURY

longer than a few days. For example, if a bird deposits excre- be determined as a maximum concentration of hydrogen in
ment on it, the effective radius will be reduced because of the a nitrogen/hydrogen mixture, and this is given by
uneven surface. This is likely even if the toroidal surface were
made of a highly polished corrosion resistant material. Cair  Clfl
Cmax ¼ (6)
There have been many other suggestions to avoid ignition (Cair  Cmoc )
made though the years to the author’s knowledge, and these
include the fitting of lightning conductors which extend above where Cmax is the maximum concentration of hydrogen in a
the flammable zone; sharp points around the edge of the nitrogen/hydrogen mixture which is non-flammable on
vent; sharp points placed some distance away horizontally dilution with air; Cair is the concentration of oxygen in air;
from the edge of the vent; and using electrostatic field Clfl is lower flammable limit concentration; Cmoc is the mini-
meters to detect the increase in potential gradient in the air. mum oxygen for combustion of the flammable constituent
The use of a lightning conductor extending above the flam- of the mixture.
mable zone is impractical because of the height required. Since the lower flammable limit for hydrogen is 4% v/v,
The use of sharp points to limit the energy of the corona and the minimum oxygen for combustion is 5% v/v, substi-
would seem to have little merit, as the magnitude of the poss- tution into equation (6) gives
ible reduction is more than offset by the potential increase in
20:8  4
field strength in thundery weather. The sharp points at some Cmax ¼ ¼ 5:26% (7)
distance away would appear to have some merit, but since 20:8  5:0
corona discharges tend to occur from more than one point, Thus a flow of inert gas of about 20 times the hydrogen rate is
it would be unlikely to prevent corona discharges occurring required. Under most applications this would be an excessive
from the edge of the vent. The use of field meters would demand, and therefore usually it is not practical to render the
have some merit, except that any emergency release vented hydrogen non-flammable under all conditions.
cannot be planned, and the shutting down of the plant for
long periods during frost, sleet and falling snow or other
weather conditions producing high electrostatic fields, Dilution with Air
would be likely to be unacceptable to a manufacturing site.
An alternative is to dilute the hydrogen with air. This would
To summarize, it would appear that despite all the best
mean dilution to below 25% of the lower flammable limit as
schemes thought up by experts, ignition of hydrogen from a
recommended by the Health and Safety Executive (HSE,
vent is inevitable at some time, and this has to be recognised
1992), but as the lower limit is 4% v/v, this means a dilution
in any design. If ignition is not acceptable, because of the
of 100-fold, and it is not possible to avoid a flammable
proximity of other buildings, personnel access and so on,
atmosphere occurring at some time as the hydrogen is
then the only alternative is dilution to a concentration which
diluted from 100%, through the flammable range to less
is non-flammable on dilution with air. Several postulated
than 1% v/v. The use of a large fan would minimize the
ignition mechanisms for generally high pressure releases
volume of the flammable mixture, but as a flammable atmos-
have been reviewed by Astbury and Hawksworth (2005).
phere could not be avoided during the dilution, such a
method cannot be recommended as being safe. Even the
DILUTION use of air-driven eductors with no moving parts cannot be
recommended as the loss of the air flow would immediately
Absolutely Inerted Mixture stop the dilution.
An absolutely inerted mixture is one which does not form a
flammable atmosphere when mixed with air in any proportion Purging
because the ratio of inert to fuel is sufficiently high (CEN,
2006). Consequently, providing sufficient inert gas is added, This involves the removal of oxygen from a vent pipe to
then the hydrogen being vented will be non-flammable ensure that the oxygen is less than the minimum oxygen
when mixed with air in any proportion. From the rectangular require for combustion in the vent itself. However, it does
flammability diagram Figure 1, as described by Bodurtha not necessarily mean that the vented hydrogen and purge–
(1980), simple geometry enables the dilution requirement to gas mixture is non-flammable when mixed with air, as the
purging is simply to ensure that the vent pipe work does
not contain a flammable mixture.

Practical work by Husa

The major work in the open literature on inert gas purging
of blow-down or flare stacks is by Husa (1964), and he pre-
sents a formula for determining the velocity at which a
purge gas has to be supplied to maintain a given oxygen
level at a certain point from the open end of a vertical
stack. The range of parameters for which the equation is
valid are for an oxygen concentration of 3 –6% v/v, stack
diameters of 0.1 –0.6 m, and purge gas molecular weights
of 2–28.
The equation shows that the oxygen concentration at any
Figure 1. Rectangular coordinate flammability diagram for hydrogen. point in the stack is an inverse function of the distance from

Trans IChemE, Part B, Process Safety and Environmental Protection, 2007, 85(B4): 289– 304
 VENTING OF LOW PRESSURE HYDROGEN GAS 293

the open end, but this is not entirely in agreement with the
experimental data for a 100 mm bore column 8.7 m long.
The equation quoted by Husa is
 0:64  n
0:022 6 28
V¼  
(H  h) X M
 ½e(0:16Di )  0:96e½0:16(Di M)  (8)

where V is the purge velocity (ft s21); H is the overall height of

stack (ft); h is the elevation of point (ft); X is the required
oxygen concentration (% v/v); n is an exponent depending
on the diameter of the pipe; M is the molecular weight of
purge gas; Di is the vent diameter (inches).
This equation is dimensionally inconsistent and empirical
in nature, but the work is based on original practical exper-
imental work carried out on full size stacks, and recognises
Figure 2. Husa’s original 1964 data re-plotted on a logarithmic scale.
that lighter-than-air gases do tend to mix with the air, Purge velocities in fts21.
whereas heavier-than-air gases tend to act more like a
piston and push the air out with minimal mixing. Husa reports
that the weather, such as rain, wind strength and direction,
and distance from the open end is logarithmic and not inver-
temperature and pressure had little measurable effect on
sely proportional as quoted by Husa. By taking the oxygen
the oxygen levels within the stacks when using permanent
concentrations at the same point in each of the three stacks
gases for the purge, but did have an effect when steam
used by Husa, the dependence of the velocity required on
was used as the purge. The accuracy of equation (8) is
diameter can be determined. The data for nitrogen is plotted
stated by Husa as being ‘. . . within 8 percent for better
in Figure 3 and gives a reasonable correlation of log oxygen
than half the data. The largest disagreement is 38 percent.
concentration against velocity, but does show that the
The standard deviation of all data is 13 percent . . .’. Clearly
250 mm pipe is quite different from the 100 mm and the
there is some scope for revising the equation to improve the
600 mm pipe, and has a lower oxygen concentration show-
accuracy.
ing that the purging is more effective at 250 mm than at the
From Husa’s 1964 equation, it can be seen that for a given
smaller and larger sizes.
stack, the term (H –h) is the distance in feet from the open
Although it appears that there is no rational explanation for
end to the point at which the defined oxygen level is required.
this, Husa’s paper does show a variation of the value of the
By considering the molecular weight of the purge gas, the
exponent n for the molecular weight term with diameter,
stack diameter and the purge velocity as constants, equation
and this curve peaks at 250 mm –see the original 1964
(8) becomes
paper. It is probable that Husa has simply fitted a smooth
  curve through the three points of 100 mm and n ¼ 1.0;
1 1 0:64
k¼ (9)
s X
[ X 0:64 ¼ ks

where k is a constant; s is the distance from the open end of

the vent; X is the required oxygen concentration, % v/v, and it
can be seen that the required oxygen concentration becomes
an inverse function of the distance from the end of the stack.
However, there must be an error, since at the situation at the
open end of the pipe, s is zero. Assuming that k is finite, then
X 20.64 must be zero, and hence X must be zero. However,
s ¼ 0 corresponds to the open end of the pipe, where the
oxygen concentration must be 21% as it is open to atmos-
phere. In view of this, it would seem that the equation of
Husa’s must be in error, and therefore, Husa’s original 1964
data was re-appraised.
Firstly, the data for the 250 mm diameter 6 m stack was
plotted as oxygen concentration at steady state against the
distance from the open end. To aid correlation with other
plots, the distance from the end was reduced to a dimension-
less number of stack diameters rather than absolute length.
Similarly, the purge velocity was reduced to the equivalent
of pipe diameters per second. Five data sets were plotted:
two data sets with nitrogen, and three with helium.
The graphs are plotted in Figure 2. From these lines, it is Figure 3. Effect of velocity of nitrogen on oxygen concentration for
clear that the relationship between oxygen concentration various pipe diameters.

Trans IChemE, Part B, Process Safety and Environmental Protection, 2007, 85(B4): 289 –304
294 ASTBURY

250 mm and n ¼ 1.4; and 600 mm and 0.83, which is hardly multi-component purge gas equation given by Husa is
conducive to accurate interpolation. It is also apparent that for    n
a gas of molecular weight 28, the exponent in the term (28/ 1 20:9 X
Qp ¼ 0:07068D3:46 ln Ci0:65
M)n does not apply as the fraction becomes unity—yet the i
Y x 1
data of Figure 3 shows that the vent diameter has an effect  
29
with a gas of molecular weight 28. The velocity dependence  exp 0:065 (13)
MW
was determined by calculating the concentration of oxygen at
a suitable purge velocity of 50 diameters per hour and plotting
Substituting for the worstPcase of hydrogen to maintain the
in Figure 3. n 0.65
stack oxygen free, the 1Ci exp[0.065(29/MW)] term in
Subsequently, a second paper was published (Husa, 1977)
equation (13) reduces to 2.566 when the molecular weight is
which used the original test work (covering stacks from 100 –
2, as opposed to the value of 1.07 for nitrogen if simply prevent-
600 mm nominal bore) and added further tests using the orig-
ing oxygen ingress when there is no hydrogen present. The
inal sizes and a 1.2 m stack. The second series of test results
term becomes 2.317 for a 50/50 v/v mixture of nitrogen and
indicated a completely different equation. The new equation
hydrogen. Therefore the final equation reduces to
was a dimensionally consistent logarithmic equation of the
form   
1 20:9
Qp ¼ 0:1814D3:46
i ln (14)
   Y x
V
x%O2 ¼ 20:9 exp  Y (10) for nitrogen being used to purge out oxygen and hydrogen.
k
However, this does have a dependence of the volumetric flow
rate on the diameter to the power of 3.46. This equates to a
where V is the purge velocity; k is the transport coefficient; Y
linear velocity of the purge gas being proportional to the diam-
is the column depth.
eter to the power of 1.46. However, this is not in accord with the
The transport coefficient, k is a function of the diameter of
original 1964 data, as the data presented in Figure 3 indicates
the stack and the properties of the purge gas. Husa’s 1977
that the 250 mm diameter pipe required significantly less vel-
paper does show a graph of k versus stack diameter, but
ocity in terms of diameters per unit time than either the
unfortunately the graph in the photocopy of the paper avail-
100 mm or the 600 mm vent. Similarly this does not equate to
able to the author had been reduced and was too small
the results of equation (14), particularly with respect to smaller
and indistinct to interpolate any values. However, it was
diameters. Husa also suggests that because the similarity of
apparent that the lines are straight when plotted using log –
density of nitrogen and air results in cross-winds affecting the
log axes. Husa then proposes a comprehensive, but again
purge, the calculated nitrogen flow should be increased by
dimensionally inconsistent, equation for purging stacks to
60% to ensure adequate purging. Consequently, although
maintain the oxygen content at 6% at a distance of 25 ft
Husa’s 1977 equation does clearly fit the data better and is of
down the stack, and this is
a more rational form, there is no justification for extrapolating
the equation to diameters less than 100 mm. Further work is
X
n
required on smaller vents of the sizes more common within
Qp ¼ 0:003528D3:46
i Vi0:65 Ki (11) the fine chemicals industry.
1

where Qp is the purge gas flowrate, ft3 h21; Di is the vent Other purging work
diameter, inches; Vi is the volume fraction of each component In the design of flares, Tan (1967) quotes a different for-
of the purge gas mixture; Ki is the transport constant for the mula, based on Husa’s 1964 results, and assuming criteria
component. P of an oxygen content of not more than 6% v/v at a distance
In order to calculate the function n1 V i0.65 Ki the individual not more than 7.62 m from the top of the stack. This assump-
component values must be summed to give the final trans- tion is based on any ignition propagating a limited distance
port coefficient. This is because the effect of mixing com- down the stack, and thereby limiting any explosion pressure
ponents is non-linear, in that the lighter and heavier to a manageable value. Since most refinery stacks are of the
components have a disproportionate effect on the buoyancy order of 45 m high, a 7.6 m length does not seem unreason-
compared to a simple mean molecular weight value. Husa able. However, in the case of typical small vents in a fine
gives a series of values for various components commonly chemicals plant, a stack of such a height would be unusual
used as purge gases. As Husa had simplified the equation and only used under exceptional circumstances. Tan quotes
to maintain the 6% oxygen at 7.62 m from the end of the
stack, a further simplification can be made by assuming CFH ¼ b M 0:565 D3i (15)
that the purge gas is nitrogen, and that the top of the
stack is under cross-wind conditions, so the equation where CFH is the purge rate in cubic feet per hour; b is a con-
becomes stant of 0.214 derived from Husa’s experimental data; M is
the molecular weight of the purge gas; Di is the stack diam-
eter in inches.
Qp ¼ B D3:46
i (12)
Because of Tan’s assumptions of an acceptable oxygen
content and maximum distance from the open end, the
where B is an constant depending on the units used for equation takes no account of variation of oxygen or distance,
the diameter and volumetric flow. In the case of any and therefore is inapplicable to short stacks less than 7.6 m
other oxygen concentration and distance, the general long. He also uses the data from Husa based on hydrogen

Trans IChemE, Part B, Process Safety and Environmental Protection, 2007, 85(B4): 289– 304
 VENTING OF LOW PRESSURE HYDROGEN GAS 295

and nitrogen, and his equation is therefore not necessarily where NRe is the Reynolds number; D is the stack diameter,
applicable for other gases or mixtures. m; Up is the purge velocity, m s21; n is the kinematic viscosity
The 1964 data collected by Husa has been analysed also of the purge gas, m2 s21.
by Panchenko (1993) who has modelled the mixing of the air From equation (17) above, an increase in diameter D is
as a bubble of air in a counter-current stream of the purge matched by a proportional increase in volumetric flow, but
gas in laminar flow. He presents an alternative equation for because the area increases by the square of the diameter,
turbulent flows, but in the circumstances prevailing on typical so the linear velocity decreases. Therefore the Reynolds
small chemical plants, the purge rate is likely to be laminar. number remains constant because of the product of diameter
The mathematically derived equation takes account of the and velocity. It is noteworthy that the Specification points out
density of both the air and the purge gas, and attempts to bal- that where the mean molecular weight is below 4.5 the appli-
ance the drag force, buoyant force, and gravity force on the cation should be discussed by the Company before any pro-
air bubble as it passes down the stack. The equation deter- posal is submitted by the Contractor, thus illustrating that
mines the buoyant force taking account of the viscosity of hydrogen vents are not normally covered by their minimum
the purge gas. However, whilst the size of the bubble purge flows stipulated in the Specification. Similarly, another
decreases as the purge gas mixes with the air, the change report (McBrien, 1988) on the venting of hydrogen states that
in viscosity of the purge gas with the admixture of air is neg- it is standard practice to purge all stacks relieving flammable
lected. The resultant analysis uses data from Husa, and gases and which are 200 mm or greater in diameter, at a vel-
therefore presents no new data, but excessively large flows ocity of 0.15 m s21, but the rate is increased during thunder-
are calculated from his formula: storms and on relief. The increased rate was not specified.
Bryce and Fryer-Taylor (1994) carried out work on measur-
!1=3
g2 Dr2 ing the length of the flammable zone within a stack, which is
4
v1 ¼ 10 D exactly what is required when determining the safety of a
v1 r21
2 3 stack. However, they used an inverted arrangement of a hea-
vier-than-air vapour falling down out of the stack for simu-
6 170:3 7 lation, using an unidentified Freonw as the heavy purge gas
6 7
6  1:147 (16)
4 y=Df½0:067=ð0:01x þ 0:053Þ 5 driving the lighter air out of the bottom of the stack. This
0:264g1 þ 0:264 may not truly reveal the behaviour of a flammable zone, as
they do not state their concentration criteria which they
where v1 is the purge gas velocity, m s21; D is the stack equate to a flammable mixture.
diameter, m; g is the acceleration due to gravity, m s22; Dr However, their analysis concludes that the practical work
is the density difference between purge gas and air, by Husa (1964) has shown that for a satisfactory purge, the
kg m23; n1 is the kinematic viscosity of purge gas stream, purge velocity, rather than purge volumetric flow, should be
m2 s21; r1 is the density of purge gas, kg m22; y is the dis- proportional to stack diameter, so that maintaining the
tance down vent for required oxygen concentration, m; x is purge gas Reynolds number constant between different
the required oxygen concentration, fractional. sized stacks will not maintain a constant oxygen concen-
Calculations substituting appropriate values into equation tration distribution. Since the Reynolds number is low, the
(16) yield excessive gas rates. Since the translation of flow regime is always laminar, and at low velocities, the grav-
paper did not have any nomenclature section, many of the itational effects are dominant rather than momentum effects.
units have had to be assumed from the text and examples Therefore scaling by constant Reynolds number is not appro-
contained within it. Overall, the paper presents little new infor- priate. Consequently the guidance based on a flow pro-
mation, and it is apparent that there is an error in the initial portional to the diameter does not seem to address the
constant of 104. Substitution of 1024 results in reasonable variation of stack diameter properly. Bryce and Fryer-Taylor
values being calculated, but this may be due to an error in (1994) state that to maintain similarity at different scales,
the translation, as only the translation was readily available. the Reynolds number is not appropriate, and they matched
However, the calculated flows for a light gas purge show the Richardson number, buoyancy ratio and velocity ratio at
values far in excess of those calculated from Husa’s 1964 experimental and full scale. Richardson number is defined as
equation. Therefore, there is little to commend this paper
for practical applications. (r0  r1 )g  d
NRi ¼ (19)
Although the above are based on practical test work, (r1 U12 )
a large chemical company Specification for Flare Stacks
(Ennis, 1998) suggests an inert gas volumetric flow based the buoyancy ratio is defined as
simply on the diameter of the stack, calculated from
(r0  r1 )
Rb ¼ (20)
Q ¼ 236 D m h 3 1
(17) r1

where Q is the inert gas purge flow rate, m3 h21; D is the and the velocity ratio is defined as
stack diameter, m.  
Based on the above equation, the Reynolds number is Uc
Vr ¼ (21)
constant for a stack with a purge flow-rate proportional to U1
the diameter. Since the Reynolds number is given by
where r0 is the density of air; r1 is the density of the purge
DU r gas; g is the acceleration due to gravity; d is the diameter
NRe ¼ (18)
n of the stack; U1 is the velocity of the purge gas; Rb is the

Trans IChemE, Part B, Process Safety and Environmental Protection, 2007, 85(B4): 289 –304
296 ASTBURY

buoyancy ratio; Vr is the velocity ratio; Uc is the velocity of the should be used. Note that equations (23) and (25) converge
cross-wind. at purge gas molecular weights of about 0.82 and 31.93.
For the same Richardson number for two different stack By substituting in the worst case buoyancy factor from
diameters, the velocity up the stack has to increase as the equation (25), taking logarithms and re-arranging, equation
square root of the increase in diameter. Thus in the case of (22) reduces to
Bryce and Fryer-Taylor, a full size stack twelve times the
p
model size would require an efflux velocity of 12, i.e., ln (21=O2 )  0:0036 D1:46
i
U¼ (26)
3.464 times as great. Therefore it would appear that the Ls  6:25½1  0:75(MW =28:96)1:5 
purge velocity should increase as the 0.5 power of the diam-
eter. Since the cross-sectional area increases as the square and the volumetric flow-rate depends on the diameter to the
of the diameter, the volumetric flow should increase by the power of 3.46. However, since the buoyancy factor is on
power of 2.5 of the diameter. Only one of the formulae the bottom line, the larger this value, the smaller velocity
above uses an exponent of 2.5 for the diameter to determine needed. Therefore the truly safe option would be to use the
the required volumetric flow. Husa (1977) uses a power of smallest gas buoyancy factor calculated by equation (25),
1.46; Tan (1967) uses an exponent of 3; Panchenko (1993) contrary to Shore’s suggestion which appears to be based
uses 4; and others use an exponent of 1 or 2 depending on on the assumption that Husa’s estimation of buoyancy is dif-
which source is consulted. These rates are summarized in ficult to determine for ill-defined mixtures and the purge rate
Table 1. is often larger than necessary.
Shore (1996) quotes a formula to determine the oxygen From the foregoing, there seems to be little evidence of any
concentration at any point down the stack for a given flow vel- reported effect of wind-speed across the top of the stack, par-
ocity, distance down the stack, diameter and molecular ticularly with Husa asserting in 1964 that the weather had no
weight. His formula appears to derive from Husa’s 1977 effect except when steam was used as the purge, and only
work as it uses the summation of individual buoyancies for purge gases with similar densities to air being affected by
the components of the purge gas, and is cross-winds being stated in the 1977 work. Husa (1977) pro-
! poses that the purge velocity should be increased by 60%
ULs Fb when cross-winds are present to take account of the ingress
O2 % ¼ 21 exp (22) of air under such conditions. Whilst the weather is likely to
0:0036 D1:46
i
have an effect at low purge rates, at high rates the mixing
where U is the gas velocity, feet s21; Ls is the distance into within the stack will tend to avoid diametral concentration gra-
stack from open end, feet; Di is the stack diameter, inches; dients, and will tend to present more vertical concentration
Fb is the adjusted the buoyancy factor. gradients, effectively as in plug flow.
The adjusted buoyancy factor Fb has three different Some work on this was carried out by Bryce and Fryer-
methods of calculation, depending on the composition of Taylor (1994). They report effects of cross-winds being gener-
the purge gas. The first, lowest factor is for single component ally to decrease the length of any flammable zone, but move
purge gases calculated using it further down the stack at low cross-wind speeds. At high
cross-wind speeds, the length of the flammable zone was
Fb ¼ exp (0:065½28:96  MW ) (23) increased, and even a flammable zone may be created
where one did not exist previously. Their variation of purge
where MW is the molecular weight. For a typical multi- gas rate showed a progressive decrease in the length of
component mixture, the buoyancy factor suggested is any flammable zone with increasing purge rate, but with a
distinct discontinuity occurring at certain rates. Further work
Xn o
Fb ¼ (volfrac)0:65 Fbi (24) indicated that vortices at the end of the stack contributed to
the problem, and whilst the centre-line flammable zone
length decreased, the flammable zone tended to move
where Fbi is the individual buoyancy factor for each com-
towards the wall, and lengthen at the same time. Thus the
ponent. Where the composition of the purge gas can vary
overall effect was to increase the volume of flammable gas
widely, or individual buoyancy factors cannot be computed
within the stack.
for each component, Shore suggests that to provide an over-
After trials at three different scales, Bryce and Fryer-Taylor
all ‘safe’ factor, the equation
concluded that a full scale purge velocity of 13 mm s21 in a
"   # 250 mm diameter stack would ensure a flammable length of
MW 1:5 less than five stack diameters. This equates to a flow of
Fb ¼ 6:25 1  0:75 (25)
28:96 2.3 m3 h21 of purge gas. The work, however, does not give
an equivalent lower flammable limit or minimum oxygen for
combustion, so cannot be used to determine a purge rate
Table 1. Exponent a for volumetric flowrate dependance on diameter. to ensure safe venting of hydrogen. Similarly, as a constant
purge gas velocity would increase the volumetric flow as
Source Exponent the square of the diameter, it would seem that the flow
McBrien (1988) 2.0 would be insufficient in stacks larger in diameter than theirs.
Bryce and Fryer-Taylor (1994) 2.5 The effect of air-ingress so far considered has been about
Ennis (1998) 1.0 the air which diffuses down from the top, open end of the
Husa (1977) 1.46 stack. Tite et al. (1989) report that air can also ingress at
Tan (1967) 3.0
Panchenko (1993) 4.0 the base of a stack, as there will be a sub-atmospheric
pressure at the bottom of the stack. This is due to the

Trans IChemE, Part B, Process Safety and Environmental Protection, 2007, 85(B4): 289– 304
 VENTING OF LOW PRESSURE HYDROGEN GAS 297

buoyancy of the gas stream in the stack, in addition to the Alternatives to Purging
reduction in pressure from the ‘chimney effect’. There have
An alternative method to purging, suggested by Burgoyne
been some incidents in flare stacks caused by air ingress
and Fearon (1986) is to utilize a perforated plate at the top of
during or after maintenance (Crawley, 1993). Therefore, the
the stack, to locally increase the velocity of the gases and
design of any drains or lutes has to take account of the poten-
thus make any flame more stable and less likely to burn
tial for air-ingress. This must also be taken into account
down inside the stack. This use of small holes relies on
where line-blinds, sample points and so on, are fitted; no
increasing the velocity, rather akin to the ‘flame stopper’
in-leakage of air can be tolerated whilst there is any hydrogen
arrester proposed by Howard (1988). Whilst this appears to
present in the stack.
be acceptable, Burgoyne suggests that practical tests be car-
Work by Hooker et al. (2001) showed that the ingress of air
ried out, and Howard proposes a safety factor of 4 on the
occurred when there was a cross-wind over the end of the
maximum flashback velocity, and that this minimum velocity
vent, with the oxygen concentration at any point down the
be used at all times to make the arrester effective. He also
vent oscillating rapidly when a cross-wind was present. How-
quotes hole sizes of up to 100 mm as being practical.
ever, these transients were not observed where only a vent
Whilst this may be effective, it still relies on a minimum gas
pipe was used. Where the vent was attached to a vessel,
velocity at all times. Howard states that the turbulent flash-
the transient effects became worse. They found that the pre-
back velocity is calculated from
sence of a large vessel attached to the vent had a profound
effect on the transients. Consequently, whilst other workers UT ¼ 0:2015 gL Dh (27)
have observed small transients close to the end of the
vent, or none at all, their experimental equipment did not where UT is the turbulent flashback velocity, m s21; gL is the
have a large volume attached. Where the ratio of total critical boundary velocity gradient (for laminar flow) below
system volume to vent pipe volume is less than 3.2, no tran- which flashback can occur, s21; Dh is the diameter of hole, m.
sients were observed by Hooker et al. (2001). However, The maximum gL for hydrogen is stated by Howard (1988),
where the ratio was high at 37 : 1, at times the entire vent quoting Grumer et al. (1956), as 10 000 s21. Equation (27)
pipe had an oxygen concentration which would support com- can be used for determining a suitable hole size, with
bustion. Hooker et al. (2001) were unable to quantify the examples shown in Table 3. For the sizes of vents previously
effect of volume, but did find that whilst the oxygen concen- considered of 150 mm, 300 mm and 450 mm, the turbulent
tration varied greatly within the vent, the oxygen concen- flashback velocities can be calculated for the stack size,
tration within the vessel did not rise rapidly. Further work on and are also given in Table 3 as an illustration. Clearly the
this aspect of the work is required. stack velocities in the lower three rows of the table are exces-
sive, but are still below sonic velocity in hydrogen which is
about 1.3  103 m s21 (Yarwood and Castle, 1961). As the
hole size increases, the flashback velocity reaches sonic vel-
ocity, and the equation becomes discontinuous. As the exit
Purging rate comparisons velocity to prevent flashback should be at least four times
A comparison of the various calculated volumetric flow- the flashback velocity, the hole size becomes limiting at prac-
rates for nitrogen is given in Table 2 for maintaining the tical gas flows. At small hole sizes, the risk of blockage is pre-
oxygen concentration at 2% in an open-ended vent or stack sent, but larger holes require higher velocities. The
at a distance of 610 mm from the open end on a 50 mm, practicality of such a method for hydrogen is somewhat
100 mm and 150 mm vent. Since some methods are specific doubtful particularly with the potential for the gas exit velocity
for 6% oxygen at 7.6 m from the open end, the last column to fall below that required to prevent flash-back.
gives these for a 150 mm vent for comparison where the
method does not allow calculation for 2% at 600 mm. The
shaded values are calculated from a suggested volumetric DISPOSAL BY BURNING
flow based on diameter, rather than for a specified oxygen
Deliberate Flaring
concentration at a specific point in the vent. All flows are in
m3 h21 to allow a comparison of the quantity required for Large quantities of hydrogen being vented could accumu-
each pipe size. late in the vicinity of the vent in calm weather, and if the

Table 2. Comparison of calculated volumetric flowrates of nitrogen to maintain a concentration of 2% oxygen 610 mm from the open end.

150 mm vent,
50 mm vent 100 mm vent 150 mm vent 6% v/v oxygen
Source (m3 h21) (m3 h21) (m3 h21) @ 7.6 m– m3 h21

Husa (1964)a 0.067 0.371 0.639 0.046

Husa (1977) 0.027 0.304 1.237 0.053
Tan (1967) 0.2
Panchenko (1993)b 0.009 0.164 0.842 0.045
Ennis (1998) 11.8 23.6 35.4
McBrien (1988) 1.11 4.45 10
Shore (1996) 0.006 0.069 0.281 0.012
a
Based on exponent n for pipe diameter set at worst-case of 1.4 for 50 mm; 1.0 for 100 mm and 1.3 for 150 mm.
b
Based on the multiplier being 1024 rather than 104 as printed.

Trans IChemE, Part B, Process Safety and Environmental Protection, 2007, 85(B4): 289 –304
298 ASTBURY

Table 3. Minimum velocities for various holes sizes to prevent Assuming that it is acceptable to allow the hydrogen to
turbulent flash-back
ignite, there may be a time limit for exposure to the thermal
Hole or Stack diameter (mm) UT (m s21) radiation and the worst case exposure needs to be con-
sidered. Kent (1968) gives some guidance on the matter,
1 2.0
and a maximum permissible burning time for the flame can
5 10.1
10 20.2 be established for radiation effects on adjacent buildings
150 302 and personnel. This will then give an indication as to how
300 604 the presence of a flame should be detected. For example if
450 907 the maximum permissible flame burning time for radiation is
1 h, it may be possible to rely on periodic visual inspection.
If the permissible exposure time is determined at less than
size of such an accumulation would cause significant over- 1 min, an automatic snuffing system may be required and a
pressure on ignition (Reider et al., 1965), then the hydrogen suitable flame detector would be required. This may be ther-
would have to be flared. Whilst this critique does not cover mocouples, passive infra-red radiation detectors, or flame
the design of flares, it will direct the decision-making process ionization detectors.
to determine whether flaring is necessary or acceptable. The If a flame does occur, the stability of the flame has to be
merits or otherwise of flare systems is discussed in a note by considered. In a similar fashion to a Bunsen burner, the
Selwa and Reid (1996). flame may be stable or unstable. At moderate flows, the
Assuming that flaring is acceptable, then the steady burn- flame will be stable. As the flow increases, the flame may
ing of the flame on the stack will need to be assessed for the tend to lift off the end of the stack, but according to Grumer
emitted radiation. Guidelines for the radiation emitted from (1970) the blow-off critical boundary velocity gradient is
flares can be estimated from several sources, such as Tan about 1028 s21, where the gradient is defined for laminar
(1967), Shore (1996), Kent (1964, 1968) and Schmidt flow as
(1985) to determine whether the radiation is excessive. If
the hydrogen to be disposed of is from a chemical reaction,

U
the possibility of contamination will have to be assessed. gB ¼ 8 (28)
Df
For example, where sodium compounds are handled, any
sodium carry-over in the gas stream will give the flame the where gB is the critical boundary velocity gradient, s21; Ū is
characteristic intense yellow ‘sodium light’, and will increase the linear velocity, m s21; Df is the diameter, m. Turbulent
the emissivity of the flame. Further information on the radi- flow is defined as
ation from flames in cross-winds is discussed by Fairweather
et al. (1992). At high velocities, flame stability can become
N 0:75
0:0395U Re
poor, and the flame can lift off and extinguish itself, particu- gB ¼ (29)
Df
larly if the flame is burning on a plain ended pipe. At the
point when the flame is extinguished, flame blow-out has where NRe is the Reynolds number. In practice, the release
occurred. If a flame retainer is fitted to the flare tip, much rates of hydrogen encountered on a typical plant will be
higher exit velocities can be tolerated. Some notes by Straitz much less than this, so flame blow-off is highly unlikely to
(1987) indicate that for methane, a maximum velocity is about occur.
Mach 0.2 to avoid flame blow-out when a flame retainer is At low velocities, the flame becomes unstable. Previous
fitted. Hydrogen is likely to have such a high flame lift-off vel- investigators have, according to Grumer (1970), treated this
ocity when a flame retainer is fitted, that noise considerations problem as one involving pre-mixed flames, in that the hydro-
may well dictate a lower exit velocity. gen and the air mix in the stack, and the flame travels back
If the radiation emitted from the flame does not affect the into the stack if the gas velocity falls below the fundamental
surrounding buildings and personnel, then the flame may burning velocity. This is the same concept as a Bunsen
be acceptable, assuming that Planning Permission has burner ‘striking back’. However Grumer (1970) proposes
either has been obtained or is not required. Note that the that the problem is one of diffusional burning in that the air
‘environmental’ aspect of the flame has to be considered, in and the hydrogen mix at the flame surface as the burning pro-
that a plain hydrogen flame is virtually invisible, but a gresses. This theory is supported by the measurements
sodium-contaminated flame would be highly visible and which showed that while helium flowed up a 450 mm diam-
may draw unwelcome attention to the stack. Similarly, the eter stack at 0.014 m3 s21, air flowed down at the same
emissivity of a pure hydrogen flame is low, but with sodium time, indicating that the light gas does not fully purge the hea-
present, the emissivity will be much higher, resulting in the vier air out of the stack at low flows. Panchenko (1993) con-
possibility of high surface temperatures on neighbouring sur- sidered this in his analysis of safe purging of stacks as
faces in direct sight of the flame. If it is desired to flare the gas discussed previously. From this, it could be concluded that
on a routine basis, then the design of the flare system should had it been hydrogen rather than helium as the purge, com-
be undertaken by specialist suppliers of the equipment. bustion could have occurred within the stack since the air
content was around 70%. This has been confirmed in prac-
tice by Grumer (1970). As the flow is reduced, the flame
Tolerated Flaring
reduces in size, and starts to cycle in height, at a frequency
Where hydrogen is vented to atmosphere and not deliber- typically about 10 –15 Hz. As the flow is reduced further,
ately flared, it may ignite due to the difficulty of avoiding the flame begins to burn inside the stack. This is termed
ignition as described above. If this is the case, then a flame ‘flame-dip’ by Grumer (1970). The experimental flows for
may be tolerated on the end of the vent occasionally. this to occur are documented for 150 mm, 300 mm and

Trans IChemE, Part B, Process Safety and Environmental Protection, 2007, 85(B4): 289– 304
 VENTING OF LOW PRESSURE HYDROGEN GAS 299

450 mm stacks, and are compared with theoretical ones, From this it would appear that there is poor correlation
showing some large discrepancies. Finally, the flame will dis- between the calculated and practical flows, but clearly, the
appear from view and burn almost completely within the experimental flame-dip velocity would always have to be
stack. It is worthy of note that Husa (1964) gives a maximum exceeded for safety. The maximum flame dip velocity
value of the molecular weight exponent (n) at a diameter of occurs at a diameter of about 300 mm, and it may be coinci-
about 250 mm, and the information of Grumer (1970) dental that the maximum for the exponent by Husa (1964)
shows a distinct maximum in the flame-dip velocity at occurs at about 250 mm diameter. It would appear that a
300 mm diameter. This is discussed later. Shore (1996) diameter of about 250 mm to 300 mm would provide the opti-
quotes an equation to estimate the burn-back velocity for typi- mum geometry to allow simultaneous down-flow of air and
cal refinery gases, but states that it under-estimates the rates up-flow of hydrogen in the same stack. Common sense
for hydrogen, due to hydrogen’s very high fundamental burn- would indicate that small pipes would not easily allow simul-
ing velocity compared to those of hydrocarbons. Shore’s taneous flows in opposite directions in a vertical section, and
dimensionally inconsistent equation quotes similarly large diameters would tend to permit good radial
mixing in a relatively short length of stack. Consequently,
½(Dl =MW )0:75 (LCV)0:5  further work is required to substantiate whether this effect is
Ub ¼ (30) real or coincidental.
1500
Having established that the minimum purge velocity should
where Ub is the approximate burn-back velocity, ft s21; Di is the exceed the flame-dip velocity, a comparison can be made
internal diameter of stack, inches; LCV is the calorific value of using the purge-rate calculations of Husa (1964, 1977). The
gas, Btu lb21; MW is the molecular weight of the purge gas. calculated purge velocities for nitrogen are based on the
Taking the calorific value of hydrogen to be 119 criteria of Grumer (1970) of maintaining the hydrogen non-
954 kJ kg21 (51571 Btu lb21), the burn-back velocities for flammable within the stack by ensuring that the hydrogen
three stack sizes can be calculated and are given in Table 4. concentration was always above 74% v/v, and hence the
These gas velocities are theoretical, and based upon the air was less than 26% v/v (i.e., an oxygen content of
rate at which the flame burns back down the stack against 5.5% v/v), at a distance of less than 610 mm from the end
the upward flow. There is no indication in Shore’s paper as of the stack. Using the equations of Husa (1964, 1977), the
to whether these are for the flame to propagate down the corresponding minimum flows are provided in Table 6.
stack or simply burn within the end of the stack. These two From Table 6, it can be seen that from practical experimen-
possibilities are described by Grumer (1970) as flame-dip tal work, flame-dip is likely to occur if either the Husa (1964)
and flame-back. From Grumer (1970) experimental work, it or the Husa (1977) velocity is directly applied with the same
would appear that the velocities calculated from Shore’s criteria as used by Grumer (1970), but flame back will not
equation are flame-dip limits rather than flame-back velocities.
The approximate limits determined by Grumer (1970) are
given in Table 5, using a twin flare stack and assuming that Table 6. Comparison between Grumer (1970), Husa (1964) and Husa
air leaks down one stack and passes up the other. The calcu- (1977) velocities to prevent flame-dip from occurring during venting of
lated values are based on the balance of air flowing down hydrogen.
one stack, and hydrogen flowing up the other. As would be Grumer
expected, there is no sharp limit between flame stability, Stack flame dip Husa (1964) Husa (1977)
flame-dip and flame-back flows. The results are also appli- diameter velocity velocity velocity
cable to single stacks in which the light gas passes up the (mm) (m s21) (m s21)a (m s21)b
centre of the stack, and the denser gas (air) flows down the 150 0.110 0.054 0.026
periphery, allowing combustion to continue within the stack. 300 0.366 0.258 0.071
450 0.238 0.454 0.128
a
Velocity calculated from equation (8) (Husa, 1964) based on oxygen
Table 4. Burn-back velocity for differing stack diameters. content of 5.5% v/v oxygen 610 mm from the end of the stack purged
with hydrogen, to correspond to the criteria laid down by Grumer
Stack diameter (mm) Burn-back velocity (m s21)
(1970).
b
150 0.105 Velocity calculated from equation (13) (Husa, 1977) based on
300 0.177 oxygen content of 5.5% v/v oxygen 610 mm from the end of the
450 0.240 stack purged with hydrogen, to correspond to the criteria laid down
by Grumer (1970).

Table 5. Grumer (1970) experimental flame dip results.

Experimental flame-back flows and

Flame-dip limits velocities

Stack diameter (mm) Predicted (kg s21) Experimental (kg s21) Velocity (m s21)a NaRe Flow (kg s21) Velocity (m s21)b

150 5.90 1024 1.59  1024 0.11 170 0.24  1024 0.017
300 13.1  1024 21.3  1024 0.37 1100 2.85  1024 0.049
450 49.9  1024 31.3  1024 0.24 1100 3.18  1024 0.024
a
Experimental flame-dip velocity and Reynolds number values, based on hydrogen flow through entire cross-section.
b
Calculated from flame-back mass flow and flame-dip limit flows.

Trans IChemE, Part B, Process Safety and Environmental Protection, 2007, 85(B4): 289 –304
300 ASTBURY

Table 7. Calculated purge rates for various oxygen criteria.

1 2 3 4 5

Stack Husa (1964) equation,

diameter Grumer O2 criteriaa Husa (1964) equationb Husa (1977) equationc Proposed criteriad
(mm) velocity (m s21) velocity, (m s21) velocity m s21) velocity (m s21)

150 0.054 1.98  1023 6.88  1023 18.9  1023

300 0.258 5.22  1023 1.90  1023 56.3  1023
450 0.454 13.6  1023 34.2  1023 76.1  1023
a
5.5% oxygen at 610 mm down the stack, based on pure hydrogen.
b
2.0% oxygen at 305 mm down the stack, based on purge gas of molecular weight 28, Husa 1964 equation.
c
2.0% oxygen at 305 mm down the stack, based on purge gas of molecular weight 28, using Husa 1977 equation, and increased by 60% in
accordance with Husa’s recommendation.
d
2% oxygen at 610 mm down the stack, based on purge gas of molecular weight of 2, using Husa 1964 equation.

occur using either flow. Therefore a larger flow than that pre- too low a purge velocity could allow air ingress down the
dicted by either Husa equation would be required to prevent vent to occur because of the buoyancy of the vented gas.
flame-dip. Under steady state hydrogen venting, the gas velocity
To compare the purging requirements, the nitrogen flow would be higher than that found necessary to prevent
can be calculated from the equation of Husa (1964), but flame-dip, and any flame would be visible at the end of the
using the worst-case exponent for the diameter as a constant stack. However, under variable flows, it would be necessary
i.e., n ¼ 1.4, a molecular weight of 28 (nitrogen), and an to maintain that gas flow, irrespective of the hydrogen flow,
oxygen content of 2% v/v oxygen but at a reduced distance to ensure that the minimum flow was always present. There-
of 305 mm from the open end of the stack where hydrogen is fore for venting of hydrogen, the minimum purge –gas
being vented. This gives velocities in the third column of requirement for the inert gas (usually nitrogen) should be cal-
Table 7, which are less than Grumer’s velocities for flame- culated according to the equation of Husa (1964), but assum-
dip. Using the same criteria as column three, but using the ing that the purge gas molecular weight is only 2. This
equation of Husa (1977) and increasing the values by 60% ensures that for the sizes of vent stacks encountered within
as suggested by Husa for cross-winds, gives the values in many fine chemicals plants (i.e., stacks typically between
column four. Taking the original equation of Husa (1964), 50 mm and 450 mm diameter), the minimum purge will
and assuming the worst case diameter exponent (n ¼ 1.4), always ensure that flame-dip does not occur.
but a molecular weight of 2 (i.e., hydrogen) and a distance
of 610 mm down the stack, yields the higher velocities Flame Extinguishing
given in the fifth column of Table 7.
If the flame appears to be stable from the criteria of Grumer
From Table 7, it would appear that the flow calculated by
the original Husa (1964) equation to give the recommended (1970), then it will be necessary to determine whether it is
values of oxygen according to Grumer (1970), are insuffi- acceptable to allow the flame to continue to burn. If it is
acceptable, then clearly the hydrogen should be vented to
ciently high to prevent the possibility of flame-dip occurring.
However, experience in the past (Sellers, 1977) has shown a properly designed flare stack which would take account
that despite frequent ignitions, no personnel injury or of the radiation, noise and so on. Where a flame does
occur on the end of the stack, the continued heating of the
damage has occurred even when hydrogen vents do ignite.
end of the stack would have to be avoided, as the heating
From the flame-back flows quoted in Table 5, it would
appear that the minimum purge gas flows quoted in Table 7 of the metal may ultimately lead to loss of physical strength,
corrosion resistance, or physical integrity.
are sufficient in 150 mm, 300 mm and 450 mm stacks to pre-
If it is not acceptable to allow it to burn continuously, then
vent flame-back, i.e., under burning conditions, the flame will
not burn inside the stack, but is likely to burn under unstable since it appears inevitable that ignition will occur at some
time, a means of extinguishing is required. There are two
conditions at the end, dipping frequently. The 610 mm depth
simple methods, which are that the hydrogen flow can be
inside the stack has been selected as this gives easily calcul-
able values based on practical experience (Husa, 1964, interrupted for a short period, or the flame can be snuffed
1977), where the flow is in excess of the minimum flame- out by the injection of an inert gas at a sufficient rate as to
be able to dilute the hydrogen to less then its lower flam-
dip velocities, and also from practical combustion work
mable limit when mixed with air.
(Grumer, 1970). Prolonged burning should not occur where
continuous venting has been designed out of a process.
Under long duration venting, such as during reactions Interruption
which release hydrogen, a higher purge gas flow would be This method relies on simply stopping the hydrogen flow
required. In the fifth column of Table 7, the proposed criteria for a short period. Clearly the minimum period is that which
are to use the Husa equation, but use the molecular weight of would allow any hydrogen within the stack to burn completely.
the purge gas as equal to that of the hydrogen. This ensures This may be difficult, particularly if flame-dip occurs, as the
that the purge velocity is sufficient to maintain inert conditions flame itself may not be visible from the ground. Any detection
irrespective of the composition of the hydrogen/nitrogen gas method such as thermocouples would have to be allowed to
mixture being vented. Under typical batch processes, the cool down, and even then, in a large diameter stack, a flame
hydrogen content of the vented gas can vary widely, and may still persist at one side of the stack. Ionization flame

Trans IChemE, Part B, Process Safety and Environmental Protection, 2007, 85(B4): 289– 304
 VENTING OF LOW PRESSURE HYDROGEN GAS 301

detectors may be more suitable, but again, they may not explosion, the noise aspect has not been covered. Expert
reliably detect a small flame. advice should be sought on the aspect of noise before pro-
The most suitable method would be to shut off the hydro- ceeding with any detailed design for the venting of hydrogen.
gen venting for sufficient time to purge the stack using the
inert gas purge. The time required to purge the stack can
be calculated from the equation CONCLUSIONS
  From the above critique, it can be seen that the determi-
V C0
t ¼ ln (31) nation of a ‘safe’ flow of nitrogen is far from simple. It would
Q Cf
appear that there are several criteria which have to be satis-
where t is the time required for purging, h; V is the volume of fied, and the minimum flow for one criterion may be exces-
the stack, m3; Q is the purge gas flow-rate (i.e., nitrogen flow), sive for another and inadequate for a third. Practical tests
m3 h21; Cf is the hydrogen content after flow purging (less will often override theory, but it is necessary to ensure that
than 4% v/v); C0 is the hydrogen content before flow purging any extrapolation from practical tests is valid. Many are not
(assume 100% v/v) (t, V and Q must be in consistent units). as can be seen from the above, where results and theories
As an example, a stack 150 mm in diameter, 10 m high, concerning similar criteria disagree.
purged with nitrogen at a velocity of 0.1 m s21, would have However, several conclusions can be drawn from the
to have the hydrogen flow interrupted for a period of about above.
5 12 min whilst the nitrogen flow is maintained. The time (1) The ideal method of disposal of hydrogen is to dilute
period may be reduced by increasing the nitrogen purge with inert gas so as to be non-flammable at all times.
rate pro-rata, so doubling the purge rate will halve the shut- This is unlikely to be practical for long-duration venting
off time. If this is not acceptable, then the flame will have to at high rates.
be extinguished by dilution. Further guidance on purging is (2) Dilution by a large volume of air could be used, but
given by CEN (2006). should be regarded as a last resort because of the
large volume required and the difficulty of dealing with
Snuffing out using a diluent the transition of passing through the flammable range
From Discharges section above, dilution would require about on dilution.
20 times the peak hydrogen flow. At a high hydrogen flow rate, (3) If the dilution is not practical, then for a continuous or
it will be necessary to use 20 times as much nitrogen. Not only long duration venting, ignition has to be regarded as
would this be expensive, but may be excessively noisy. inevitable.
An alternative requiring careful design would be to use (4) If a vent can ignite, and the duration or intensity of any
steam as the diluent. The advantage is that on a plant, it is flame is unacceptable, then it may be necessary to
often available in large quantities and at low cost. The disad- design a suitable snuffing-out system to extinguish the
vantage is that it is hot, will contain condensate initially, and flame.
produces a highly visible plume. The same constraints (5) Dispersion of hydrogen is rarely a problem because of
apply regarding noise. its buoyancy, but in calm weather it can accumulate
If the use of steam is contemplated, then it will be necess- and form a flammable gas cloud near the vent.
ary to ensure that the design takes into account: (6) If ignition of such a cloud would give rise to an unaccep-
table over-pressure, then the vent must be relocated to
(1) means to ensure adequate removal of the condensate
a more remote area or the release must be prevented
produced, without allowing the ingress of air into the
from accumulating in a flammable cloud.
base of the stack;
(7) Any stack venting hydrogen will operate with a sub-
(2) the thermal expansion of the stack, to ensure that any
atmospheric pressure at the base, and suitable draining
guy-wires or supporting structures are not over-stressed;
arrangements will have to be designed to allow draining
(3) the effect of wind-speed and air temperature on the con-
of condensate and rainwater without the admission of
densation of steam within the stack (the ‘wind-chill’ factor)
air. Where line-blinds, other isolation devices or
to ensure that an adequate volumetric flow of steam
sample points are used, the admission of air into the
leaves the top of the stack;
stack must be avoided whilst they are brought into
(4) the potentially corrosive effect of the hot, wet environment
operation.
within the stack;
(8) Bursting disc tail-pipes require inert gas purging to pre-
(5) the potential for the downward ingress of air when the
vent flammable concentrations of hydrogen forming
steam condenses within the stack at the end of the
within the pipe after a release.
purge period;
(9) Where a non-balanced relief valve discharges directly to
(6) the potential for re-ignition to occur from electrostatic
atmosphere, no inert gas purge is required providing
charging of the droplets of condensate, particularly if
that the discharge pipe-work and valve body can with-
the steam flow is interrupted or reduced below that
stand a detonation, typically in the order of 30–40 bar.
required to dilute the hydrogen.
A balanced-bellows valve will usually be unable to with-
stand such a high pressure on the discharge side, and
hence the tail-pipe would have to be protected by an
inert gas purge.
NOISE
(10) Inert gas purges are required to prevent any flame that
Since this critique is more concerned with the venting of becomes established on the end of the vent from burn-
hydrogen and the prevention of ignition, flash back or ing down inside the stack. A suitable rate for this is

Trans IChemE, Part B, Process Safety and Environmental Protection, 2007, 85(B4): 289 –304
302 ASTBURY

difficult to decide upon, but the indications are that a Summary Flowchart for Venting Hydrogen
flow calculated from Husa’s 1964 equation (8) which is The steps below give an indication of an approach to a
equal to or above the minima obtained by any experi- design procedure to ensure that hydrogen is vented safely.
mental or theoretical work to avoid the flame propagat- Each individual case will have to be considered on its own
ing down the stack would be required. This rate is well merits, so there is no ‘global solution’ which will be applicable
in excess of that required simply to maintain a low in every case. Similarly, the solution already in use for one
oxygen content and can be calculated using the follow- vent should not be simply repeated on the next vent—due
ing assumptions: consideration to any change of circumstances should be
(a) the oxygen content is 2.0% v/v; made. To facilitate the appropriate decisions, a small decision
(b) the maximum distance from the end of the pipe is tree is given in Figure 4 to allow an appropriate design to be
610 mm; established.
(c) the exponent for the molecular weight term n is 1.4 if
the vent is below 100 mm diameter; (1) Determine the maximum rate and duration of the hydro-
(d) the molecular weight of the purge gas is taken to be gen release, and determine the release profile.
2, even though the purge gas is nitrogen with a mol- (2) If the release is only a small quantity, then a simple
ecular weight of 28. purge is all that is required.
(11) Intermittent venting of small quantities of hydrogen (as in (3) If the release is for substantial periods, where the maxi-
de-pressurizing a vessel) can be safely carried out using mum release rate is small, and can be easily diluted with
a continuous inert gas purge up the vent-pipe. Rates of 20 times the flow of nitrogen, then this is the best
flow calculated from Husa’s 1964 equation (5) would solution.
appear to be satisfactory for this duty, using the following (4) If the required dilution flow cannot be provided, then it
assumptions: will have to be assumed that the hydrogen will ignite
(a) the maximum oxygen content should be 2.0% v/v; at some time, and a suitable inert gas purge rate
(b) the maximum distance from the end of the pipe decided upon.
should be taken to be 305 mm; (5) Assuming that ignition will occur, then determine the
(c) the molecular weight of the purge gas should be estimated flame size and radiation. Take into account
assumed to be 28 for nitrogen. the emissivity of the flame which may be high if contami-
nated with high emissivity substances, such as carbon

Figure 4. Decision tree for establishing appropriate venting regime for hydrogen.

Trans IChemE, Part B, Process Safety and Environmental Protection, 2007, 85(B4): 289– 304
 VENTING OF LOW PRESSURE HYDROGEN GAS 303

dioxide, sodium ions and so on. If this is not acceptable, Clfl lower flammable limit concentration
Ci concentration of ith component in the purge gas, % v/v
then the vent may have to be relocated further from the
C0 hydrogen content before flow purging, % v/v
critical exposure area. Cmax maximum concentration which is non-flammable on
(6) Assuming that the exposure to the flame is acceptable, dilution with air
then determine the dispersion and dilution under varying CMOC minimum oxygen for combustion of the flammable
atmospheric conditions. This should be sufficient to CFH purge rate, ft3 h21
D stack diameter, m
ensure that the flammable envelope does not include d diameter of the stack
any buildings, structures or places where personnel Df diameter, ft
may pass. Cude (1974) and commercial CFD computer Dh diameter of hole
programs may assist in this respect. A discharge vel- Di vent diameter (inches)
ocity will have to be assumed, but since hydrogen has Ec threshold field, kV cm21
Fb adjusted buoyancy factor
a low molecular weight, buoyancy effects will outweigh Fbi individual buoyancy factor
any momentum effects. Once the maximum extent of g acceleration due to gravity
the flammable zone is established, determine the peak gB critical boundary velocity gradient, s21
over-pressure on ignition. gL critical boundary velocity gradient (for laminar flow)
below which flashback can occur, s21
(7) If the peak over-pressure is too high, then it will be H overall height of stack, ft
necessary to relocate the vent, or deliberately ignite it. h elevation of point, ft
If permanent ignition is decided on, then the vent i number of component in gas mixture
should be designed as a flare stack, and the manufac- k transport coefficient, ft2 s21
turer of the stack should be able to provide an accepta- Ki transport constant for the i th
component in the purge gas
ble design. Ls distance into stack from open end
(8) If the peak over-pressure is acceptable, then the vent is LCV calorific value of gas, Btu lb21
satisfactory, but the exit velocity of the total gas flow M molecular weight of purge gas
(i.e., hydrogen and inert gas purge to prevent air MW molecular weight
n an exponent depending on the diameter of the pipe
entry) should be determined. It should not exceed a Nma Mach number
Mach number of 0.2 because of the potential for noise. NRe Reynolds number
At 291 K and atmospheric pressure, the velocity of NRi Richardson number
sound in hydrogen is about 1300 m s21 and in air is Q generalized purge gas flowrate (i.e., nitrogen flow)
about 342 m s21. The speed of sound can be estimated Qp purge gas flowrate, ft3 h21
p R universal gas constant
for any perfect gas using the equation Us ¼ (gRT/MW) r radius, cm
quoted by Perry et al. (1997). Rb buoyancy ratio
(9) Once the maximum flow is established and the diameter T absolute temperature
of the vent has been calculated, then the minimum inert t time required for purging
U gas velocity, feet s21
gas purge to prevent air ingress and flame-dip should be U1 generalized velocity of the purge gas
determined. This may have an effect on the dispersion Ū linear velocity, ft s21
as it affects the mean molecular weight of the gas, Ub approximate burn-back velocity, ft s21
and the purge rate should be recalculated to ensure Uc velocity of the cross wind
Up purge velocity, m s21
that the proposed inert gas purge still gives acceptable
Us velocity of sound in the gas, ft s21
dispersion. UT turbulent flashback velocity
(10) Assuming that all is acceptable, determine the maxi- V purge velocity, ft s21
mum acceptable duration of burning. From this a suit- n1 kinematic viscosity of purge gas stream, m2 s21
able means of extinguishing should be determined. Vi volume fraction of each component of the purge gas
mixture
Ideally this should be interrupting the hydrogen flow Vr velocity ratio
temporarily to extinguish the flame. If this is not poss- Vs generalized volume of the stack
ible, then a large flow of inert gas will be required to Vs potential of the electrode, V
dilute the hydrogen to less than 5% v/v. Note that this X oxygen concentration, fractional
will again affect the dispersion, the stack efflux velocity, x required oxygen concentration, percent
Y column depth, ft
and the Mach number, as the change in mean molecular y distance down vent for required oxygen
weight will alter the velocity of sound in the emitted concentration, m
gases.
(11) If sufficient inert gas cannot be provided in a short-
duration high flow purge, then steam can be used, bear-
ing in mind the points discussed above regarding steam Greek symbols
purging. Dr density difference between purge gas and air, kg m23
g ratio of specific heats at constant pressure to constant
volume
n kinematic viscosity of the purge gas m2 s21
NOMENCLATURE r0 density of air, kg m23
r1 density of the purge gas, kg m23
a exponent for flow dependency on diameter v1 purge gas velocity, m s21
B arbitrary constant
b a constant of 0.214 derived from Husa’s experimental
data
Where no units are quoted, the number is either dimension-
Cair conc of oxygen in air (¼21% v/v) less or the units are irrelevant providing consistent units are
Cf hydrogen content after flow purging, % v/v used throughout the equation.

Trans IChemE, Part B, Process Safety and Environmental Protection, 2007, 85(B4): 289 –304
304 ASTBURY

REFERENCES Husa, H.W., 1977, Purging requirements of large diameter stacks,

presented at the Fire/Safety Engineering Subcommittee of
Anon, 1972, Imperial Chemical Industries Petrochemicals Division the American Petroleum Institute, Fall 1977 Meeting, 13–15
Safety Newsletter 72/3, p. 2. September, Holiday Inn (Union Square), San Francisco, California,
Astbury, G.R. and Hawksworth, S.J., 2005, Spontaneous ignition of USA.
hydrogen leaks: A review of postulated mechanisms, Proceedings Institute of Energy, 2005, Model Code of Safe Practice—Part 15:
of the HySafe International Conference on Hydrogen Safety, 8 –10 Area Classification Code for Installations Handling Flammable
September, Pisa, Italy. Fluids, 3rd edition.
Barnett, H.C. and Hibbard, R.R., 1957, Basic considerations in Institution of Chemical Engineers, 1994, Explosions in the Process
the combustion of hydrocarbon fuels in air, National Advisory Industries, Major Hazards Monograph, Second edition.
Committee for Aeronautics, Report No. 1300. International Standards Organisation (ISO), 2000, Technical Report
Bodurtha, F.T. 1980, Industrial Explosion Protection and Prevention, 15916:2000, Basic considerations for the safety of hydrogen
(McGraw Hill, New York, USA). systems.
Bradburn, F. and McBrien, D., 1983, ICI Organics Division Internal Kent, G.R., 1964, Practical design of flare stacks, Hydrocarbon
Report TVR194 (not available in the public domain). Processing and Petroleum Refiner, 43(8): 121 –125.
Bryce, S.G. and Fryer-Taylor, R.E.J., 1994, Reducing the purge in Kent, G.R., 1968, Find radiation effects of flares, Hydrocarbon
hydrocarbon vent stacks, J Loss Prev Process Ind, 7: 249–255. Processing, 47(6): 119– 130.
BSI, 2003, BS EN 60079 10:2003 Electrical apparatus for explosive Long, V.D., 1963, Estimation of the extent of hazard areas round a
gas atmospheres—Part 10: Classification of hazardous areas. vent, IChemE Second Symposium on Chemical Process Hazards,
Burgoyne, J.H. and Fearon, R., 1986, The flash-back hazard in 6– 14.
stacks discharging flammable gases, Institution of Chemical McBrien, D., 1988, Private communication to the author.
Engineers Symposium Series No. 97. Ministry of Technology, 1968, Prevention of gas ignition at vent stack
Burrows, D.A. and Hobbs, P.V., 1970, Electrical charges on snow outlets, Technical Note Serial No. 218, Subject Code 52, Techlink
particles, J Geophysical Research, 75(24): 4499– 4505. Unit, UK (information from NASA Brief B67-10098).
Cadwallader, L.C. and Herring, J.S., 1999, Safety issues with hydro- Panchenko, V.I., 1993, Determining the safe purge rates for flare
gen as a vehicle fuel, Idaho National Engineering and Environ- tubes, Chem and Petr Eng, 29(5/6): 205–209 [translated from Khi-
mental Laboratory Report INEEL/EXT-99-00522, available at micheskoe i Neftyanoe Mashinostroenie, 5(5–7): 1993].
http://energy.inel.gov/fossil/hydrogen/pdf/h2safetyreport.pdf. Perry, R.H., Green, D.W. and Maoloney, J.O., 1997, Chemical
Camp, P.R., 1976, Charge, morphology and pH of natural snow, J Engineers’ Handbook, 7th edition, 6–22 (McGraw-Hill,
Geophysical Research, 81(9): 1589– 1592. New York, USA).
CEN, 2006, CEN/TR 15281:2006, Guidance on inerting for the pre- Reider, R., Otway, H.J. and Knight, H.T., 1965, An unconfined, large-
vention of explosions. volume hydrogen/air explosion, Pyrodynamics, 2: 249– 261
Crawley, F.K., 1993, Flare and drain systems—the twin Achilles heel of Schmitt, D.A., 1985, Radiation from a hydrogen flare in a crosswind,
process plants? Loss Prevention Bulletin Number 111, 1–4, IChemE. 23rd Aerospace Science Meeting, AIAA, New York, USA.
Cross, J.A., 1987, Electrostatics: Principles, Problems and Appli- Sellers, J.G., 1977, Confidential ICI Mond Division Internal Report
cations, 369– 370 (Adam Hilger, Bristol, UK). MD18641 (not available in the public domain).
Cude, A.L., 1974, Dispersion of gases vented to atmosphere from Selwa, R. and Reid, A., 1996, Vent & flare systems—the problems,
relief valves, The Chemical Engineer, October: 629– 636. The Chemical Engineer, 7 March, 24–26.
Ennis, A., 1998, Private communication to the author. Shore, D., 1996, Making the flare safe, J Loss Prev Process Ind,
Fairweather, M., Jones, W.P. and Lindstedt, R.P., 1992, Predictions of 9: 363– 381.
radiative transfer from a turbulent reacting jet in a cross-wind, Straitz, J. F., 1987, Flare technology safety, Chem Eng Prog,
Combustion and Flame, 89: 45– 63. 83(7): 53– 62.
Grumer, J., 1970, Hydrogen flare stack diffusion flame low and high Swain, M.R. and Swain, M.N., 1992, A comparison of H2, CH4, and
flow instabilities, burning rates, dilution limits, temperatures, and C3H8 fuel leakage in residential settings, Int J Hydrogen Energy,
wind effects, U.S. Bureau of Mines Report of Investigations 17: 807–815.
No. 7457. Swain, M.R. and Swain, M.N., 1996, Passive ventilation systems for
Grumer, J., Harris, M.E. and Rowe, V.R., 1956, Fundamental flash- the safe use of hydrogen, Int J Hydrogen Energy, 21: 823– 835.
back, blowoff, and yellow-tip limits of fuel gas-air mixtures, U.S. Tan, S.H., 1967, Flare system design simplified, Hydrocarbon
Bureau of Mines Report of Investigations No. 5225. Processing, 46: 172–176.
Health and Safety Commission, 1979, Advisory Committee on Major Tite, J.P., Greening, K. and Sutton, P., 1989, Explosion hazard of air
Hazards, second report (HMSO). ingress into flare stacks, Chem Eng Res Des, 67: 373 –380.
Health and Safety Executive (HSE), 1992, Flame arresters and Webber, D.M., Jones, S.J. and Tickle, G.A., 1997, Dispersion of
explosion reliefs, Health & Safety Booklet HS(G)11, 4th very buoyant plumes, IChemE Symposium Series No. 141,
impression. 89–100.
Hooker, P., Astbury, G.R. and Fauré, G., 2001, Air ingress to nitrogen Yarwood, T.M. and Castle, F., 1961, Physical and Mathematical
inerted chutes, IChemE Symposium Series No. 148, 541–554. Tables, 2nd edition (Macmillan & Co., London, UK).
Howard, W.B., 1988, The velocity flame stopper, Loss Prevention
Bulletin No. 079, 15– 21, IChemE.
Husa, H.W., 1964, How to compute safe purge rates, Hydrocarbon The manuscript was received 27 August 2006 and accepted for
Processing and Petroleum Refiner, 43: 179–182. publication after revision 1 December 2006.

Trans IChemE, Part B, Process Safety and Environmental Protection, 2007, 85(B4): 289– 304