You are on page 1of 10

Journal of Loss Prevention in the Process Industries 24 (2011) 612e621

Contents lists available at ScienceDirect

Journal of Loss Prevention in the Process Industries


journal homepage: www.elsevier.com/locate/jlp

Ventilation theory and dispersion modelling applied


to hazardous area classification
D.M. Webber, M.J. Ivings*, R.C. Santon
Health and Safety Laboratory, Harpur Hill, Buxton SK17 9JN, UK

a r t i c l e i n f o a b s t r a c t

Article history: Critical formulae given in the current Explosive Atmospheres Hazardous Area Classification Standard IEC
Received 15 November 2010 60079-10-1 (2008) [BS EN 60079-10-1, 2009] to determine the expected gas cloud volume which is used
Received in revised form to determine area classification do not have any scientific justification. The standard does allow the
6 April 2011
alternative use of Computational Fluid Dynamics (CFD) methods, which serve to compound the concern
Accepted 8 April 2011
with these formulae: the predicted volume of the gas cloud from CFD models being several orders of
magnitude smaller than that given by the formulae in question. To resolve such major discrepancies,
Keywords:
replacement of the current formulae with a scientifically validated approach is proposed. Integral models
Area classification
Jets
of dispersion and ventilation have been used routinely for many years in the analysis of major hazards in
Flammable gas the chemical industry. This paper presents an adaptation of these models to determine the expected
Integral model volume of a gas cloud arising from a release of gas from a pressurised source. A very simple integral jet
Ventilation model is presented for outdoor dispersion, extended to the case of indoor dispersion, from which the
volume of the gas cloud is derived. The single free parameter, an entrainment coefficient, is fixed by
comparison with data on a free jet, and then predictions of the model are compared with CFD calcula-
tions (which themselves have been validated against experimental data) for dispersion within an
enclosed volume. The results of this simple integral model are seen to agree very well with the CFD
predictions. The methodology presented here is therefore proposed as a scientifically validated approach
to Hazardous Area Classification.
Crown Copyright Ó 2011 Published by Elsevier Ltd. All rights reserved.

1. Introduction relevant International Standard IEC 60079-10-1, 2008 (published in


the UK as BS EN 60079-10-1, 2009) (hereinafter “the standard”) as:
The control of sources of ignition by the use of specially pro-
Zone 0 e a place in which an explosive gas atmosphere is
tected equipment in areas where flammable gases or vapours may
present continuously or for long periods or frequently. (Exam-
arise has been a fundamental safety measure for many years.
ples e inside a closed vessel, near the liquid surface in an open
Following the availability of specially protected electrical (flame-
vessel.)
proof) equipment for use in mines in the early 20th century, it was
Zone 1 e a place in which an explosive atmosphere is likely to
soon adapted for use in surface chemical industries. The classifi-
occur in normal operation occasionally. (Examples e Sample
cation of hazardous areas into “divisions” (later called “zones”) was
points, relief valves, drainage points.)
introduced in the 1960s. It was recognised that the highest level of
Zone 2 - a place in which an explosive atmosphere is not likely to
protection is required where the risk of a release is highest, and that
occur in normal operation but, if it does occur, will persist for
lower levels of protection could be used where the risk of release is
a short time only. (Examples e near flanges, pipe fittings, valve
lower without prejudice to overall safety.
stems, pump glands.)
Thus hazardous areas are classified into zones based on the
expected frequency of occurrence and the expected duration of an The ATEX 137 Workplace Directive (1999/92/EC) has been
explosive gas atmosphere. The zones are currently defined in the implemented in the UK as DSEAR, the Dangerous Substances and
Explosive Atmospheres Regulations (2002) and by similar regula-
tions in other EU member states. These regulations require
Hazardous Area Classification (HAC) to be carried out where there
* Corresponding author. Tel.: þ44 1298 218133; fax: þ44 1298 218840. may be a risk of explosion due to the presence of flammable
E-mail address: matthew.ivings@hsl.gov.uk (M.J. Ivings). substances in the form of gases, vapours, mist or dust. To ensure

0950-4230/$ e see front matter Crown Copyright Ó 2011 Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.jlp.2011.04.002
D.M. Webber et al. / Journal of Loss Prevention in the Process Industries 24 (2011) 612e621 613

safe operation, any equipment (electrical or non-electrical) used in other methods that values of Vz can result which are up to 3 orders
a classified hazardous area falls within the scope of the regulations of magnitude smaller than those arising from using the formulae in
and must be suitable for use in the respective zone. Accurate and the standard.
justifiable area classification is therefore not just a technical safety It is clearly inappropriate that a standard should give the option
requirement but also, in the EU, a legal requirement. Specifically, to of using two methodologies where the results between these show
ensure safety and legal compliance, the methodology used for HAC such wide and inconsistent discrepancy. New methodologies are
needs to be both proven and reliable. Without this, the level of therefore required for those who do not have access to, or choose
safety realised will be uncertain and the cost of providing the not to use CFD for the estimation of Vz indoors and outdoors based
specially protected equipment for use in the Hazardous Areas on sound scientific theory to address this concern.
determined, to be unjustified, or potentially worse, insufficient. In fact the methodology required is not entirely ‘new’. Integral
This can only be avoided if there is a proper understanding of models of dispersion and ventilation have already been developed
the physical processes leading to the hazard, in terms of both from the 1970s onwards for the analysis of ‘major industrial
experiment and theory. If a scientific analysis can be done which hazards’ following the explosion involving cyclo-hexane at Flix-
leads to a ‘best estimate’ of the hazard, then it can be examined borough in 1974; the 1976 Seveso dioxin incident in Italy (leading to
carefully to reveal the circumstances under which it might be an the EU’s so-called ‘Seveso directive’); and the toxic cloud from the
underestimate, and appropriate allowance made in order to Union Carbide plant in Bhopal (1984). These incidents and others
produce a conservative estimate. If there is no scientific basis, then have spawned an immense amount of experimental and theoretical
any safety factors introduced into an estimate (however that is research into the behaviour of gas clouds, which can be adapted
done) remain in the realm of conjecture and the overall level of quite straightforwardly to provide simple formulae for the esti-
safety realised must be uncertain. mation of Vz. This is the purpose of the current work.
The methods presented here focus on the science and making As noted above, the simple Vz concept is a useful and practical
a best estimate, but one particularly clear danger will become one. It must be noted, though, that the standard stresses that Vz is
apparent: the hazard is very sensitively dependent on the hole size only to be used as an assessment of the ‘degree of ventilation’ and
leading to a release, and an underestimate of that size will poten- not as having any relation to the location of the hazardous zone. For
tially lead to severe non-conservatism. We therefore emphasise example the standard makes no reference to how the potentially
throughout that the results presented here, if they are used for area explosive mixture is distributed (although it has in fact sometimes
classification, must be interpreted with care if a conservative esti- been used with spurious geometric considerations). In the process
mate is to be guaranteed. of adapting fluid models to the estimation of Vz some indications of
The existing standard classifies hazardous areas on the basis of the distribution of the cloud will emerge, in some cases at least, but
the size of an expected “hazardous volume” Vz of gas. If this volume we emphasise that these should be interpreted with care. And Vz
is less than 0.1 m3, it is suggested that, if ignited, the cloud would will be shown to be dependent not only on the ‘degree of ventila-
produce such small overpressure and thermal effects that it may be tion’ but also, and often more significantly, on the source of gas.
regarded as insignificant. In this case the area classification is The prime objective here is to provide a scientifically-based
designated as of “Negligible Extent” (NE) and no further action is estimation of Vz, specifically in the case of a hazard from pressur-
required. Therefore protected equipment and controls over sources ised vessels or pipe-work, which can be used in the manner defined
of ignition are not required. Accepting that there is empirical by the standard, but with a much greater degree of confidence.
evidence for this classification (Ivings et al., 2008), this paper will
focus on methods for estimating Vz. 2. The current standard e a summary
The hazardous volume Vz is defined as a volume within which
the volume average concentration of the cloud is equal to a certain For indoor situations, in a ‘room’ of volume V0, the current
threshold, depending on the Zone classification. For Zone 2 standard defines a volume Vk by
conventionally the threshold is taken as 50% of the lower explosive
limit (LEL); for Zones 0 and 1 it is taken as 25% of LEL. This paper qmin
Vk ¼ V (1)
will denote this threshold as the ‘critical threshold concentration’ q1 0
(Ccrit) and not explore further which of these values might be
a better choice in what circumstances, but it should be borne in where q1 is the ventilation volume flux (volume of air per unit time
mind that these are the sort of figures of interest e typically mole flowing into and out of the room) and qmin is the value of q1 which
fractions of no more than a few percent. results in a room-average concentration equal to the critical
Other aspects of IEC 60079-10-1 are more confused however. threshold concentration of interest (Ccrit). Vz is found by multiplying
The parameter Vz is used in the standard to differentiate between this volume by a factor f, ranging from 1 to 5 according to whether
‘high’, ‘medium’, and ‘low’ ventilation in an enclosed room. This ventilation is considered ‘ideal’ or ‘impeded’ to some degree. As noted
misses the essential point that Vz may depend also on the source of in the introduction, no justification at all is provided for this formula.
gas, and in many cases it may be more sensitively dependent on the Out of doors the ‘room’ is considered to be a cube of 15 m side,
source than on the ventilation of the room. This misunderstanding ‘ventilated’ by a breeze of 0.5 m/s entering and leaving by opposite
is taken to its logical conclusion in the case of outdoor jet releases, faces of the cube e thus prescribing q1 ¼ 112.5 m3/s. Curiously (in
which in reality can dilute essentially independently of the state of view of the immense amount of knowledge gained about the
subject in the last 40 years) the standard makes essentially no
the atmosphere, but where Vz in the standard is determined by the
‘ventilation’ of a fictitious cube with sides of 15 m by an imagined mention of the concept of ‘dispersion’.
air flow of 0.5 m/s. The introduction of these spurious dimensions,
and neglect of genuine factors determining dispersion, produce 3. Ventilation theory
a completely arbitrary result for Vz.
As an alternative to the formulae it presents, the standard Consider a room of volume V0 containing a source of hazardous
expressly allows the use of Computational Fluid Dynamics (CFD) to gas released at a rate (volume per unit time) qs. Let the rate of air
calculate an estimate of Vz. But it is now clear (Benintendi, 2010; inflow from ventilation be q0 and the total outflow be q1, as shown
Gant & Ivings, 2005; Ivings et al., 2008) from the use of CFD and in Fig. 1.
614 D.M. Webber et al. / Journal of Loss Prevention in the Process Industries 24 (2011) 612e621

If we require the room-average concentration to remain lower


than the critical threshold concentration Ccrit, then we must
demand a minimum degree of ventilation:
   
Cs qs  Cs
q1 > hq*min  qs hqmin (8)
Ccrit 3 Ccrit
where the last inequality applies when 3  1. The final value qmin is
the minimum required ventilation flux (in the absence of any
knowledge of how well distributed the ventilation flow is within
the room), and is exactly the minimum ventilation rate mentioned
in the standard IEC 60079-10-1 (2008). It must be emphasised that
it is based entirely on the existence of solid walls, and on
constraints on room-average concentration. It has nothing to do
with the distribution of gas in the room, nor with any Vz which
might be significantly smaller than the room volume, and abso-
lutely nothing to do with outdoor dispersion.
Fig. 1. Volumetric fluxes in a ventilated room.
Ventilation considerations alone tell us very little about the
hazardous volume itself, but one further consideration links
Assuming that the concentration of gas at the outflow aperture ventilation theory with the standard, which defines a volume
is C1, a balance of fluxes yields:
qmin
Vk h (9)
Volume : q1 ¼ q0 þ qs (2) n
based on the minimum required ventilation flux and the air
dC exchange frequency. The standard then asserts that the resultant
Contaminant : V0 b ¼ Cs qs  C1 q1 (3) hazardous volume Vz is equal to f$Vk without any form of deriva-
dt
tion. The factor f is defined to have a value between 1 and 5, and is
where Cs is the concentration at the source and Cb is the back-
described as the “efficiency of ventilation” (presumably meaning
ground concentration.
“inefficiency” as the hazard increases with f) due to “impeded air
The solution of these equations requires an assumption about
flow”.
how well-mixed the air is within the room away from the jet. If this
The physical significance of these volumes can be clarified using
is optimally efficient, then the background concentration will be
the following thought experiment employing Maxwell’s Demon2 to
essentially uniform, and we will have C1 ¼ Cb. More generally let us
take all of the contaminant gas in the room and mix it with just
assume
enough air to give a uniform concentration Ccrit. This leaves the
C1 ¼ 3Cb (4) room divided into two zones: one containing contaminant mixed
with air to a uniform concentration Ccrit, and the other containing
with a constant 3, which we shall denote the ‘efficiency of back- pure air e see Fig. 2.
ground mixing’ as 3 ¼ 1 defines a well-mixed room volume. If the This must be done so that the room-average concentration
ventilation flux enters and leaves in a part of the room distant from remains as Ce as predicted by ventilation theory (including the
the gas source, then we may expect 3 < 1: the room is ventilated but circulation efficiency factor 3). It is readily found that the volume of
the background mixing is such that the air is not having the optimal the mixture in these circumstances is:
effect in diluting the jet.1
These equations are solved straightforwardly to give Ce Cs q s Cs qs V0 q V
V ¼ V ¼ ¼ min ¼ k (10)
h i Ccrit 0 Ccrit 3q1 0 Ccrit 3q1 3n 3
Cb ðtÞ ¼ Ce 1  e3nt (5)
If we associate the standard’s “ventilation inefficiency” f with
the inverse of our “efficiency of background mixing” 3, then this is
where the equilibrium room-average concentration Ce (achieved
just the hazardous volume asserted by the standard! Interestingly,
asymptotically at large time) is
under the assumptions of this derivation, the volume Vk/3 provides
Cs qs qs an absolute upper bound on Vz: there is no rearrangement of
Ce hCb ðt/NÞ ¼ ¼ Cs (6) molecules in the room which can make Vz bigger than this. On the
3q1 3ðqs þ q0 Þ
other hand, we had to employ Maxwell’s Demon to create this
and cloud; in reality we expect any dispersion mechanism to produce
a hazardous volume very much less than this value.
q1
n¼ (7)
V0
is the air change rate (frequency) from ventilation. Note that the
asymptotic room-average concentration depends on the source and 2
Conceived as a thought experiment by James Clerk Maxwell and subsequently
ventilation fluxes; the air change rate determines how rapidly it is
named by Lord Kelvin, the demon is a magical being who can open or close a door
achieved. in a wall when a molecule approaches from one side or the other depending simply
on whether or not he wishes to let it through. The demon can therefore arrange
that molecules with a given property congregate in one part of the room and other
molecules congregate elsewhere, separating, for example, faster molecules from
1
The term ‘efficiency’ must be taken somewhat loosely as a jet close to, and slower (hot gas from cold), or in this case contaminant from air. The question of
directed at, the outlet, may result in a value 3 > 1. This situation is not of primary how the demon thus apparently manages to reduce the entropy of the gas while
interest, however, as it would be a fortuitous circumstance producing an optimistic doing no work, has been recurring theme of interest ever since; we shall not review
estimate of any hazard. the subject here but simply take it for granted that this being is mythical.
D.M. Webber et al. / Journal of Loss Prevention in the Process Industries 24 (2011) 612e621 615

there is a background concentration of hazardous gas. Therefore


define also

Cb [ML3] the background concentration of flammable gas


rb [ML3] the background density

The case of outdoor dispersion in pure ambient air is recoverable


in the limit (Cb, rb) / (0, ra). For indoor jets a non-zero value of Cb
will be taken from the ventilation theory presented above, defining
our model, QUADVENT, of jet dispersion in a ventilated room.

4.1. Subsonic jets and plumes

Consider initially subsonic jets and passive plumes. The simple


1-dimensional steady-state flow equations are derived by consid-
ering conservation of momentum, mass, and contaminant gas.
Fig. 2. A cloud of uniform concentration Ccrit and volume Vk.

  Jet Plume
Momentum : d ru2 A (11)
In fact if we consider how Vk was derived above, then one thing ¼ 0 u ¼ ua
dz
becomes clear immediately: there has been absolutely no consid-
eration of how the gas actually disperses within the room. The
Jet Plume
above discussion of the factor f in the standard highlights the fact Mass : dðruAÞ dðruAÞ
that the standard combines both the rate at which air enters the ¼ ð2prÞrb uE ¼ ð2prÞrb uE
dz dz
room (ventilation), the way that air may be distributed around the
(12)
room (background mixing), and the way the gas mixes with air
within the room (dispersion) under the single description ‘venti-
lation’3, and this underlies much of its confusion. In reality it is Jet Plume
dispersion, and not just ventilation, which determines the extent of Contaminant : dðCuAÞ dðCuAÞ
¼ ð2prÞuE Cb ¼ ð2prÞuE Cb
a hazardous cloud, and so it is to dispersion theory that we now dz dz
turn. (13)
Where the cross-section A and radius r are related by
4. Dispersion of jets and plumes

A ¼ pr 2 (14)
In most cases of interest the hazard is a gas in a pressurised
vessel or pipe-work, and so this will be the primary consideration Looking first at the mass equation e which is of the same form
of this paper. A small breach will result in a jet. Gas jets have been for jet and plume: this describes a mass flux in direction z which
studied since the 1960s and in great detail since the 1980s. Their increases downstream owing to entrainment of ambient air of
behaviour is well understood. Many computerised safety analysis density rb with some entrainment velocity uE(z), through the
program suites contain gas jet modules, but the standard requires perimeter of the plume. (Again it is worth remembering that no
a simple analytic formula for Vz. To this effect we present the lowest absolute boundary need exist: these concepts are valid in the
common denominator of such models, which we denote QUADJET.4 context of self-similar profiles.)
For comparison purposes a simple model of a passive plume The contaminant equation describes the contaminant concen-
(denoted QUADPLUME) is presented in the same framework. tration (mass per unit volume) in the same way.
The notation will be as follows: The momentum equation is a genuine conservation of
momentum flux for the jet case (the jet is considered to be in calm
z [L] downstream distance coordinate air), but in the plume case simply a statement of advection with
A(z) [L2] the cross-sectional area (as a function of down- a wind travelling at constant speed ua.
stream distance) Treating the hazardous gas and air as ideal gases at constant
r(z) [L] the jet/plume radius (as a function of downstream temperature and pressure with molecular weights Mg and Ma, the
distance) density and concentration of the mixture satisfy
u(z) [Lt1] the downstream flow velocity (as a function of
downstream distance) r  ra ¼ lC (15)
C(z) [ML3] the concentration of flammable gas (as a function of
where l ¼ 1(Ma/Mg).
downstream distance)
Finally the models are completed by a sub-model for entrain-
r(z) [ML3] the plume/jet density (as a function of downstream
ment. Under the assumption of fully turbulent flow (high Reynolds
distance)
number) the entrainment velocity is restricted by dimensional
considerations. In the case of a jet, the turbulence velocity in the jet
where the square brackets indicate dimensions. It is also of
which powers entrainment is simply proportional to the jet
interest to consider a jet within a large indoor volume in which
velocity u. In the case of a plume in the open, the turbulence level is
proportional to the ambient air speed. Therefore the entrainment
3
velocity should be proportional to u in both cases. The simplest sub-
This leads further to the bizarre situation in which outdoor dispersion is
considered by the standard to be a form of ventilation.
model defined by uE ¼ au was originally proposed by Morton,
4
The name QUADJET is intended as a reminder of the fact that the model is to be Taylor, and Turner (1956) for buoyant plumes and is usually
considered as the absolutely simplest one possible: a QUick And Dirty JET model. referred to under their names. But a priori a dependence on the
616 D.M. Webber et al. / Journal of Loss Prevention in the Process Industries 24 (2011) 612e621

dimensionless ratio r/rb cannot be ruled out, and another popular Following this procedure yields a relationship which is very
model is the model of Ricou and Spalding (1961) simple in the limit where the critical threshold concentration is
rffiffiffiffiffi small compared with the source concentration (CH/Cs<<1):
r
uE ¼ a u (16)
rb
Jet Plume
pffiffiffi
     3=2
which we shall adopt here. At the level addressed here, the choice 9prs3 rb 3=2 Cs  Cb 3 pr3 3 Cs  Cb
VZ z VZ z s
of model is not expected to make an enormous difference, and this 16a rs Ccrit  Cb a Ccrit  Cb
one leads to significantly simpler analytic formulae. Note however (27)
that the entrainment coefficient a must be determined empirically,
and that there is no reason to expect it to be the same for plumes For a jet out of doors there is zero background concentration and
and jets as the turbulence structure in the two cases may be very the background density is that of pure air. In this case therefore:
different. (In fact for plumes in the open, it may depend on atmo-
Jet Plume
pffiffiffi
spheric stability.)     3=2
Equations (11)e(16) define the jet and plume models, in a way 9prs3 ra 3=2 Cs 3 pr3 3 Cs (28)
VZ z VZ z s
which makes the physical assumptions clear. The solutions are 16a rs Ccrit a Ccrit
straightforward. Denoting conditions at the source with subscript
‘s’, they are
For the jet: 4.2. Properties of jets and plumes
us
Jet Velocity : u ¼ (17) Two properties of these equations merit immediate comment.
½1 þ bðz=rs Þ
Firstly, the hazardous volume is proportional to the cube of the
source radius, multiplied by a dimensionless function of the ratio of
ðCs  Cb Þ source and critical threshold concentrations. This follows simply
Jet Concentration : C  Cb ¼ (18)
½1 þ mðz=rs Þ from dimensional analysis. This dependence on source size is
sensitive (for example underestimating a breach radius by a factor
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
of 2 will underestimate the hazardous volume by a factor of 8), and
Jet Radius : r ¼ rs ½1 þ bðz=rs Þ½1 þ mðz=rs Þ (19)
therefore the choice of credible breach size will be crucial to esti-
with constants mating a hazardous volume. A particular point of concern is that at
high overpressures (typically higher than around 10 barg) an
rffiffiffiffiffi rffiffiffiffiffi initially small puncture may propagate into a larger breach. In this
rb r
bh2a mh2a s (20) case, rather than trying to estimate a realistic hole size, it may be
rs rb better, given this sensitivity, simply to estimate Vz as being large
For the passive plume: enough to pose a significant hazard.
Secondly, there is no dependence of Vz on the velocity of the
Plume Velocity : u ¼ ua (21) released gas in the case of the jet, or the ambient air speed in the
case of a passive plume. This contrasts starkly with the current
standard, where results (for subsonic jets) depend both on source
ðCs Cb Þ radius and source velocity through the mass release rate. The origin
Plume Concentration : C Cb ¼
(22)
rh i
of the lack of dependence on velocity is very simple. The entrain-
1þ s ð1þbðz=rs ÞÞ2 1
rb ment rate is determined by the turbulence velocity scale in the jet
or plume. And that scale is proportional to the jet velocity or the
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

rs h i
ambient air velocity respectively. So the faster the gas moves, the
2
Plumer Radius : r ¼ rs 1þ ð1 þ bðz=rs ÞÞ 1 (23) faster it entrains, and if one is interested purely in spatial concepts
rb
(like hazard range or hazard volume) then the velocity just cancels
with out of the problem. A consequence is that there is no dependence
rffiffiffiffiffi on storage pressure. This last conclusion is, however, only expected
rb b
bha h (24) for subsonic jets; the extension of this analysis to sonic jets will be
rs 2
given below.
Consider the point zH downstream where the concentration has Thirdly, if the hazardous concentration of interest, Ccrit, is small,
decreased to a value CH. The volume of the cloud upstream of that then the jet has a larger hazardous volume than the passive plume e
point is given the same source radius. This is because the jet velocity dies off
indefinitely downstream (under the assumption of no ambient
ZzH flow) and so also, therefore, does entrainment.
VH ¼ dz$AðzÞ (25) It is also interesting to consider how an ambient flow would
0 affect a jet release. Near the source, where the jet velocity is much
higher than the ambient velocity, the jet equations may be taken as
and the volume averaged concentration hCiH within this volume is
a good approximation to reality. Further downstream, as the jet
given by
velocity decays to the ambient velocity, then plume behaviour may
ZzH set in, and the result for Vz may be between that of the two models
1 given above. Significantly in the jet, both concentration and
hCiH ¼ dz$CðzÞAðzÞ (26)
VH velocity decay inversely as downstream distance, and so if the
0
source velocity obeys
The results of these two integrals give VH and hCiH as functions
of zH, and eliminating zH will therefore give VH as a function of hCiH . us Cs
> (29)
Setting hCiH ¼ Ccrit will yield VZ as defined in the standard. ua Ccrit
D.M. Webber et al. / Journal of Loss Prevention in the Process Industries 24 (2011) 612e621 617

then one may assume that the hazardous zone lies entirely in the  ð g Þ
region where jet behaviour is dominant. In any event, the jet 2 g1
Bh / 1:89 (31)
formula, will be expected to give a conservative result for any given gþ1 g/1:4
source radius.
Finally, it must be acknowledged that one assumption under- Results are not expected to depend significantly on the ratio of
lies all of this analysis: that of turbulent flow. This is expected for specific heats, g, and we shall adopt the typical value of 1.4 in
all but the very slowest releases in a very poorly ventilated room, order to provide specific numerical results. In this case choked
but if such conditions prevail, then care should be taken in flow is expected in releases from overpressures of 0.89 bar or
interpreting this work. The defining quantity is the Reynolds higher.
number Re ¼ usrs/n or Re ¼ uars/n where n here is the kinematic For unchoked flow, resulting in subsonic jets, Britter (1994)
viscosity of air (approx 105 m2/s). A 1 mbar overpressure is reports that an isobaric jet model of the form presented above
expected to result in a release of order 10 m/s release, and from may be used with source radius rs given directly by the aperture
a 1 mm radius hole therefore Re w 1000. For this or larger or faster size.
releases the assumption of turbulent flow is acceptable. For
a passive plume from a 1 mm hole on a 0.5 m/s ambient flow then rs ¼ r0 (32)
Re w 50 and turbulent flow is not expected (but this implies
The source velocity is approximated by
a release rate of only 1.6  106 m3/s and so a hazardous volume
may take some time to build up).
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u2 0 13
u    g1
4.3. Sonic and subsonic jets pffiffiffiffiffiffiffiffiffiffiffiu6 2 B
g1 p g C7
us ¼ gRTa u t41  g  1 @B g  p A5
a
The above analysis is for subsonic jets. A sonic jet results from
choked flow in the aperture. This occurs when the pressure vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
0 1ffi
u
behind the breach exceeds a certain threshold. Immediately u    g1
pffiffiffiffiffiffiffiffiffiffiffiu 2 B p g C
outside the breach there is a region where the pressure in the jet h gRTa u t g1 @ p 1A ð33Þ
a
is higher than atmospheric and, because of this, it is generally
considered that no air is entrained in this zone. Atmospheric
pressure is achieved a few hole diameters from the source, and
In which R is the universal gas constant divided by the molecular
this point can be considered a pseudo-source for an isobaric jet
weight of the gas, and we have replaced the expected temperature
which behaves as described above. In this case rs in the above
in the orifice by ambient temperature Ta (as the likely error
formulae is not the actual hole radius but the radius of the
involved in this is only of order 10%).
pseudo-source, which may be somewhat larger, as illustrated
For choked flow resulting in sonic jets, Bitter shows that liter-
schematically in Fig. 3.
ature yields a plethora of estimates of rs/r0 all arising through
The high pressure zone is relatively small but complicated: the
a desire to find a simplified understanding of the very complicated
depressurisation is accompanied by Mach discs and barrel shocks,
shock zone. A variety of authors, including Birch, Brown, Dodson,
as shown, for example, by Ewan and Moodie (1986) in a shadow-
and Swaffield (1984), Birch, Hughes, and Swaffield (1987),
graph picture. Because the zone is generally small compared with
Britter’s (1994) own recommendation, Ewan andpMoodie ffiffiffiffiffiffiffiffiffiffiffiffiffiffi (1986),
the overall jet volume of interest, the main objective in studying
find that this ratio increases proportionally with ðp=pa Þ at large
this zone is to find the ratio of the pseudo-source radius to the
values of p. The underlying reason is simple enough. As the
actual hole radius, rs/r0. The dependence of the jet volume on the
pressure drops from p to pa, then the density drops also, and if,
storage pressure enters via this quantity. The other question of
broadly speaking, the temperature change is bounded, the density
prime interest is, of course, whether a sonic, or subsonic, jet is
drops by a factor comparable with the pressure drop. If the
expected in any given source scenario.
velocity in the region from the nozzle to the pseudo-source is
Both of these questions are addressed in an extensive review by
constrained by the speed of sound, then a conserved mass flux
Britter (1994) which gives the following guidance.
means that the jet cross-sectional area must grow to balance the
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
The important parameter is the ratio of storage pressure to
density drop, yielding rs =r0 w ðp=pa Þ. The coefficient of pro-
ambient pressure p/pa. For punctured vessels the consensus is that
portionality, and the deviation from this asymptotic behaviour at
unchoked flow resulting in a subsonic jet will occur if
smaller values of p/pa, differ from model to model, none of which
p are entirely compelling. In view of this, a pragmatic approach is
B (30)
pa simply to set

with sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi
rs p
¼ 1þK B (34)
r0 pa

with a phenomenological constant K and Britter’s (1994) review


suggests that K z 0.5 is a reasonable value. This formula behaves
according to the general consensus at large p/pa and is consistent
with the unchoked flow result when p/pa ¼ B.
The flow velocity at the pseudo-source is taken to be the speed
of sound:
pffiffiffiffiffiffiffiffiffiffiffi
us ¼ gRTa (35)
Fig. 3. Radial expansion of a jet in the depressurisation zone immediately outside the
release aperture. again giving continuity when p/pa ¼ B.
618 D.M. Webber et al. / Journal of Loss Prevention in the Process Industries 24 (2011) 612e621

In both choked and unchoked cases, the (pseudo-) source 6.3. Walls and obstructions
density and concentration are
pa The above model neglects to consider walls and other obstruc-
Cs ¼ r s ¼ (36) tions. Some qualitative arguments about their effect can be made
RTa
however.
5. The entrainment coefficient Small obstructions, such as pipe-work, should not change the
result for Vz by a large factor. If a jet encounters a small obstacle,
The value of the entrainment coefficient, a, must be found with then it may be deflected and/or envelop the obstacle. Turbulence
recourse to experiment. This is most conveniently done by generated in the wake will entrain more air, broadening the jet and
comparing the velocity decay of subsonic air jets, and thus avoiding slowing it down. But the turbulent energy responsible comes ulti-
complications due to density factors and the optimum value of the mately from the jet, and whilst the shape of the hazardous volume
pseudo-source radius. Birch et al. (1987) give very clear data for may be different, we expect only a small difference in the magni-
such jets showing the centreline velocity decay of the form tude of the hazardous volume. In general, slowing the jet, and
increasing its radius, will not have a large effect on Vz as long as the
u=us z4:89  2  rs =z (37) process responsible primarily involves entrainment of air.
The meaning of ‘small’ in the above paragraph should be
for z more than a few nozzle diameters downstream, from which quantified. It means small on the scale of the jet radius at the point
we extract b ¼ 0.10 and hence where the obstacle is encountered. As an example, consider that
Ivings et al. (2008) have performed CFD calculations which found
az0:05 (38)
that a cubic obstacle with the same width as the source diameter,
and placed just over 5 source diameters from the release point in
6. Summary of predictions of Vz the path of a jet, increased Vz by a factor of 2. A spherical obstacle in
similar circumstances had less effect. But a factor of 2 in Vz is
6.1. Outdoors equivalent to increasing the hole radius (or the linear dimension of
the hazardous zone) by only 25%. Quantifying the effect of the
The hazardous volume for outdoor jet releases is predicted by obstacle in terms of linear dimensions, therefore puts it into
Equation (28) in which rs is the aperture radius for unchoked perspective as a ‘small’ effect. In fact regarding the larger Vz
releases (see criterion (30)) and a pseudo-source radius (Equation essentially as a 25% increase in linear dimension invites the
(34)) for choked releases. For a pure source it is convenient to conjecture that it is related to the fact that the linear size of the
define the vol/vol concentration (or mole fraction) obstacle is about 30% of what the free jet diameter would have been
at the point in question.
xcrit ¼ Ccrit =Cs (39) More significant may be what happens when a jet encounters
a larger obstacle or a wall5. For this case it is pertinent to invert the
so that for the jet release question and ask how an obstruction could increase Vz significantly.
    The dependence on rs3 again gives useful insight here. Sonic jets
9prs3 ra 3=2 1 3 give a Vz which increases with pressure because rs is larger at higher
VZ ¼ (40)
16a rs xcrit pressure: the effective source is larger, but is still pure gas. An
obstruction which could significantly increase the effective source
6.2. Indoors radius, without diluting the jet, would also lead to a larger value of
Vz. It is possible to imagine exotic ways in which this might happen.
The hazardous volume for indoor jet releases is predicted by For example, suppose that the jet enters a container through a small
Equation (27) with the same source radius considerations. In this hole, is partly blocked by internal obstructions, and leaves through
case there is also a dependence on the room background concen- a much larger hole. If the container prevents any air from entering,
tration, which we can take from the considerations of ventilation then it acts as a damping reservoir of gas, admitting a fast narrow
theory. Taking into account the bounding value of Vz as the room jet and letting out a wider slower one, thus increasing the value of
volume, V0, we can estimate Vz. Of course in practice one does not expect there to be such an
(     ) obstacle in exactly the right place, but consider now a jet directed
9prs3 rb 3=2 1  xb 3 from very short range into a corner of the room. One can imagine
¼ min ; V0 ; xb < xcrit
VZ 16a rs xcrit  xb (41) that a certain amount of almost pure gas can gather in the corner
¼ V0 ; xb  xcrit itself, and that the reflected jet (or jets) may be slower and broader,
and lead to a larger Vz.
where the background vol/vol concentration is defined by consid- Viewed from this perspective (in terms of linear dimensions) the
erations of ventilation and background mixing: process does not have to be very dramatic to have a significant effect:
qs a factor 5 in the effective radius, for example, would result in a factor
xb ¼ (42) of 125 in Vz. CFD provides an appropriate framework in which to
3q1
improve upon these qualitative arguments, as will be seen below.
For any given size of hole, this provides an interesting rela- First however, let us note that the release point would have to be
tionship between the hazardous volume and the room background very close to the wall or corner (and directed at it) to produce
concentration. For low background concentration the dependence a significant effect. While the room background concentration
is weak: the hazardous volume depends essentially entirely on remains low, Equation (18) tells us that a jet centreline
the properties of the jet. For higher background concentrations, the
hazardous volume depends more strongly on this: in this case the
hazardous volume depends both on the source and on the degree of 5
A model of a wall jet would be useful in this case but is outside the scope of the
ventilation, and may also be sensitive to the efficiency of back- current work. For example, the larger Vz found by Ivings et al. (2008) for a jet very
ground mixing 3. This is discussed further below. close to, and parallel to, a wall should be accessible to integral modelling.
D.M. Webber et al. / Journal of Loss Prevention in the Process Industries 24 (2011) 612e621 619

Fig. 4. Some of the volumes studied (not to scale), denoted “Very small” (1 m3), “Small” (8 m3), “Medium” (44.7 m3), and “Large” (400 m3), showing obstacles, and the Vz predicted
by CFD from some of the jets considered. In each case air enters through openings on the right and leaves on the left.

a 100.0000
b 100.0000

10.0000
10.0000

1.0000
1.0000
Vz (m3)

Vz (m3)

0.1000
0.1000

0.0100 0.0100

0.0010 0.0010

0.0001 0.0001
1 10 100 1000 1 10 100 1000
average gas concentration at outlets (%LEL) average gas concentration at outlets (%LEL)

c 100.0000 d 1.0000

10.0000

1.0000 0.1000
Vz (m3)

Vz (m3)

0.1000

0.0100 0.0100

0.0010

0.0001 0.0010
1 10 100 1000 1 10 100 1000
average gas concentration at outlets (%LEL) average gas concentration at outlets (%LEL)

Fig. 5. Predictions from CFD (filled symbols) and the integral model QUADVENT (hollow symbols) for Vz plotted against the concentration at the ventilation outflow. Points
corresponding with different jet releases and ventilation rates are included in each figure. Circles e very small enclosure, Diamonds e small, Squares e medium, Triangles e Large.
620 D.M. Webber et al. / Journal of Loss Prevention in the Process Industries 24 (2011) 612e621

concentration of 1% (say) is achieved when z is about 500*rs. For On the other hand, gas jet models of the kind presented here
rs ¼ 1 mm, that would mean that the jet must be released within have been used routinely for the assessment of major hazards for
about 0.5 m of the wall for the wall to have a significant effect on Vz. many years and, despite their essential simplicity, their scientific
credibility is well established. It is therefore natural to apply them
7. Comparison of the integral model with CFD results to hazardous area classification of pressurised flammable gases.
Whilst the basic integral models presented here cannot be
The current standard explicitly allows the use of CFD modelling considered innovative, two relatively new features do emerge. First,
to estimate Vz, though it makes no mention of the kind of integral an analytic formula for the hazardous volume Vz defined by the
modelling presented above. It is therefore appropriate to compare standard can be derived from the model. Secondly, a standard jet
our results with those of CFD computations. model can be extended to cover jet dispersion in a room in which
there is a non-zero background concentration of flammable gas.
7.1. Unobstructed jets in a room This results in a hazardous volume which depends not only on the
degree of ventilation of the room but also on the source.
Ivings et al. (2008) have performed CFD computations in rooms The jet model itself has one free parameter, the entrainment
of different sizes with air entering through one or two apertures in coefficient, which has been fixed by comparison with the data of
one wall, and leaving through one or two in the opposite wall. In Birch et al. (1987). The extension to indoor jet dispersion has been
some cases there was an ‘obstacle’ within the room in the form of tested here, by comparing with the CFD results of Ivings et al. (2008),
either a large rectangular parallelepiped, or a large cylinder, at with the predetermined entrainment coefficient, plotting hazardous
assorted positions. Jet releases were studied varying the position of volume against the concentration in the flow leaving the ventilated
the source, the direction of the jet, and the release rate. They also room. The agreement is striking. Given that Ivings et al. (2008) also
studied a variety of these configurations experimentally. The good compared the CFD approach with experimental results, this must be
agreement found lends confidence to the CFD results in the whole considered as strong support also for the integral model approach.
set of scenarios studied. The analysis presented here also suggests that more may be
As a validation exercise on the simple integral model QUADVENT, gained by going further down this path e specifically in regard to
let us look at a subset of the scenarii where the jet did not imme- obstructed jets, aerosol jets, and buoyancy. The detailed CFD work of
diately impinge on a wall. These are described as “very small”, Ivings et al. (2008) shows that larger values of Vz can result if the jet
“small”, “medium” and “large”, according to the room size, and in the emerges very close to an obstacle of some kind. We have discussed
last two cases, the room contained a large obstacle, as shown in Fig. 4. how the linear dimension of a small obstacle compared with the
In each case the jet is fairly central in the room and ventilation diameter of (what would otherwise be) a free jet at that point
inlets and outlets are on opposite walls with the jet in between. There suggests an increased linear cloud dimension (on which Vz depends
is no reason to assume poor circulation within the room and so we as the cube) and this may be a useful topic for a further investigation.
take 3 ¼ 1. Graphs of predictions of Vz using QUADVENT and CFD Ivings et al. (2008) achieved a significant increase in Vz by directing
against the concentration at the outlet are shown in Fig. 5 for a number the jet into a very confined space constrained not only by the three
of different jet sources and ventilation rates as documented by Ivings surfaces meeting at the corner of the room, but also by a large solid
et al. (2008). Especially in view of the fact that no parameters have structure very nearby. The analysis presented here points to the idea
been tuned, the agreement is excellent. It could be improved further that it may, in future, be possible to allow for such awkward config-
with only a slight increase in the QUADVENT’s entrainment coeffi- urations with relatively simple rules of thumb. To that aim we also
cient, but the agreement (and, as it turns out, very slight conservatism) recommend further work on modelling wall jets in a ventilated room.
using the predetermined entrainment coefficient more than The focus of this work has been on pressurised gas releases, but
adequately demonstrates the validity of the integral model approach. models for two-phase jets (e.g. from LPG storage) are used routinely
in major hazards analysis, and these too should be adaptable to
7.2. Jets with a greater degree of obstruction Explosive Atmospheres Hazardous Area Classification.
In the case of pressurised releases, the jet is usually moving so
Ivings et al. (2008) also consider, as their most severely fast that buoyancy is not an important factor. However buoyant
obstructed case, a release directed at a wall and within the small releases (see e.g. Fanneløp & Webber, 2003; Morton et al., 1956) can
space, 50 cm wide, between the rectangular object and the wall. be considered in the same framework: there are cases where the
This results in a significantly larger value of Vz as would be expected turbulence generated from buoyant plume rise can dominate the
from the discussion presented with the summary of integral model background turbulence in the room, and the methods equivalent to
predictions. It is interesting that this might be approachable in those developed here for jets can also be used to estimate Vz in this
integral model terms, based on the relative dimensions of the free case.
jet and the space into which it is directed, but following this Our suggested formulae for Vz are therefore not the last word,
conjecture is beyond the scope of the current work. but do provide a valid approach where the hazard is defined by the
release of flammable gas from a pressurised source. Analyses no
8. Conclusions and recommendations more complicated than the one presented here for gas jets can be
straightforwardly applied to other hazards, and would result in
We have presented a simple integral model of a gas jet in the formulae applicable to a wider range of hazard scenarios.
context of Hazardous Area Classification (HAC) relevant to This publication and the work it describes were funded by the
the standard IEC 60079-10-1, 2008. The standard is centred on the Health and Safety Executive (HSE). Its contents, including any
estimation of a ‘hazardous volume’, the magnitude of which is used opinions and/or conclusions expressed, are those of the authors
to classify the degree of ventilation and hence in turn the HAC. alone and do not necessarily reflect HSE policy.
However, the formulae for Vz given in the current standard are not
scientifically based, and in general cannot be expected to produce
References
results that reflect reality. Furthermore Vz isn’t even a measure of
the degree of ventilation as it depends both on the ventilation and Benintendi, R. (2010). Turbulent jet modelling for hazardous area classification.
on the source of gas. Journal of Loss Prevention in the Process Industries, 23, 373e378.
D.M. Webber et al. / Journal of Loss Prevention in the Process Industries 24 (2011) 612e621 621

Birch, A. D., Brown, D. R., Dodson, M. G., & Swaffield, F. (1984). The structure and Ewan, B. C. R., & Moodie, K. (1986). Structure and velocity measurements in
concentration decay of high pressure jets of natural gas. Combustion Science and underexpanded jets. Combustion Science and Technology, 45, 275e288.
Technology, 36, 249e261. Fanneløp, T. K., & Webber, D. M. (2003). On buoyant plumes rising from area sources
Birch, A. D., Hughes, D. J., & Swaffield, F. (1987). Velocity decay of high pressure jets. in a calm environment. Journal of Fluid Mechanics, 497, 319e334.
Combustion Science and Technology, 52, 161e171. Gant, S. E., & Ivings, M. J. (2005). CFD modelling of low pressure jets for area classi-
Britter R. E. (1994). Dispersion of two-phase flashing releases e FLADIS field fication. Report HSL/2005/11. Buxton: Health and Safety Laboratory.
experiment. Cambridge Environmental Research Consultants report FM89/2 to IEC 60079-10-1. (2008). Electrical apparatus for explosive gas atmospheres. Part 10.
the Commission of the European Communities DG XIII. Classification of hazardous areas.
Bs EN 60079-10-1. (2009). Electrical apparatus for explosive gas atmospheres. Part 10. Ivings, M. J., Clarke, S., Gant, S. E., Fletcher, B., Heather, A., Pocock, D. J., et al. (2008).
Classification of hazardous areas. Area classification for secondary releases from low pressure natural gas systems,
DSEAR. (2002). The dangerous substances and explosive atmospheres regulations HSE Research Report RR630. HSE Books.
2002. UK Statutory Instrument 2002 No. 2776. Morton, B. R., Taylor, G. I., & Turner, J. S. (1956). Turbulent gravitational convection from
EU. (1999). Directive 1999/92/EC of the European Parliament and of the Council of maintained and instantaneous sources. Proceedings of the Royal Society A, 234,1e23.
16 December 1999 on minimum requirements for improving the safety and Ricou, F. P., & Spalding, D. R. (1961). Measurements of entrainment by axisym-
health protection of workers potentially at risk from explosive atmospheres. metrical turbulent jets. Journal of Fluid Mechanics, 8, 21e32.

You might also like