Академический Документы
Профессиональный Документы
Культура Документы
o
o
1
for initial conditions C = 0 when t = 0.
The maximum concentration (volume fraction) that can occur throughout the well mixed
volume is given by:
C
m
V
max
=
-
o
To estimate some values for the leak rate than would give rise to flammable concentrations
in an enclosure we can take some typical values:
For a dense gas:
V = 1000 m
3
corresponding to a building with typical dimension
10 m.
= 2 kg/m
3
density of the flammable gas e.g Cat A.
o = 12/3600 = 0.0033 s
-1
an adequate ventilation rate of 12 changes per
hour:
| | C
max
. , . e 0 02 010
a typical flammable limit range, [LFL,UFL] by
volume fraction
.
31
from which we find that | | m
-
e 013 0 66 . , . kg/s. A factor 2 variation on these end range values
would be a reasonable uncertainty to apply. The time taken to reach 90% of maximum
concentration is ~700 s.
Thus, to round figures, if the gas is well mixed, a release of ~ 0.1 kg/s could give rise to
potentially flammable conditions through an adequately ventilated building, of volume 1,000
m
3
, on a time scale of ~700 s. This would be an extremely hazardous condition. Larger
releases might lead to potential flammable regions outside of the building because the
material leaving would be above the lower flammability limit for the gas. Certainly a release
as large as ~ I kg/s, capable of filling the building to the upper flammability limit should not
be tolerated.
The basic premise of the ventilation calculation is that the volume should be well mixed.
Mixedness will not occur through the ventilation flow itself as the implied ventilation wind
speed is very small, of order 33 mm/s assuming a typical building length-scale of 10 m.
Some other stirring mechanism is needed.
If we consider the pressurised releases treated in section 3, then a release rate of ~ 0.1 kg/s
for a category A or B fluid could arise for a hole size as small as 2 mm for pressures below
50 bar. The hazard radius for an unconfined release of this magnitude is ~ 5 m which is half
the characteristic length scale of the example building and the order of magnitude of the free
path of a centrally placed release. The hazard radius is also the distance over which the
unconfined jet entrains enough air to dilute the mass flow to the LFL. This entrainment is
comparable to the circulation in the building and it is thus reasonable to propose that the
volume of air passing through the building can be well mixed by a pressurised release.
For a lighter than air gas:
We can rework the above example for the category G(i) and G(ii) gases. We have:
V = 1000 m
3
corresponding to a building with typical dimension
10 m.
= 0.78, 0.29 kg/m
3
corresponding to a category G(i) and G(ii) gases
respectively.
o = 12/3600 = 0.0033 s
-1
an adequate ventilation rate of 12 changes per
hour:
C
LFL
= 0.046, 0.04 m
3
/m
3
LFL for category G(i) and G(ii) gases respectively.
.
giving m
-
= 012 . , 0.04 kg/s for category G(i) and G(ii) fluids respectively.
From these simple arguments we can deduce that quite small pressurised releases of the
type used to underpin the hazard radius recommendations for unconfined releases and of
size ~ 0.1 kg/s could result in the establishment of a flammable mixture within a 1000 m
3
building having a ventilation rate of 12 air changes per hour for Category A and G(i) fluids,
and a somewhat lower value of ~ 0.04 kg/s for category G(ii) fluid.
32
External Hazards
The implication of this is that there may be an efflux of flammable material from the buildings
for releases of order 01
1000
.
V
kg/s for category A and G(i); 0 04
1000
.
V
kg/s for category G(ii)
gases where V is the volume adequately ventilated. The fate of this material needs to be
assessed.
Flow and mixing around buildings
Flow and dispersion are strongly influenced by the building shape and orientation to the
wind. Buildings are classified as bluff rather than streamlined bodies and this greatly
complicates the description of the flow. Consequently here is a large body of work reporting
the study of flow and dispersion in building wakes.(e.g. Castro & Robins, 1977; Hosker,
1979; Britter, Hunt & Puttock, 1976; Snyder and Lawson, 1994).
The main feature of a bluff rather than a streamlined body is that the wind flow cannot pass
smoothly around the body but instead separates from the upwind edges of the body. This
forms a region along the sides and behind the body in which pressure is reduced. The
pressure difference across this region causes the separated flow from each edge and side of
the body to curve towards each other and eventually intercept to form a recirculation region
behind the body in which the mean flow is actually reversing as fluid is returned towards the
low pressure regions. Depending on the shape of the body this region can extend back up
the building sides to the building front. The flow within the recirculation region is highly time
varying and unsteady.
It was originally thought that the recirculation region was a closed "bubble" bounded by a
separation streamline and that material was transported in and out of the region by turbulent
mixing across this boundary. This remains a useful simplification of the flow but it is now
known that transfer of material in and out of the recirculation region is not limited to
turbulent transport across the boundary. Material is also advected into the region along
entering mean streamlines and advected out via vertical spiral vortices or through
entrainment by horseshoe vortices. The entire recirculation region may even collapse
intermittently causing all the contents to be flushed downstream.
There is no simple and accurate way of assessing the fate of material released into the wake
of a building. Dispersion models in regulatory use (ADMS, ISC, AERMOD) recognise that
significant mixing takes place in a building wake and this can reduce concentrations
downwind of the building by up to an order of magnitude compared with an unconfined
release. The models do not attempt to describe events in the near wake and treat this as a
well-stirred region of constant concentration i.e. any material leaving the building is predicted
to undergo a step change in concentration.
The models should not be used closer to the building than 3-5 times the extent of the
recirculation zone. Typically the extent of the recirculation zone for squat shaped buildings
scales as the building height and for tall thin buildings it scales as the building width. For a
building of characteristic dimension 10 m this imposes an a region of modelling uncertainty
of 30-50 m. This is of greater extent than the present guidelines for the external hazard
radii.
An alternative to regulatory models is to use experimentation or Computational Fluid
Dynamics to explore some release scenarios. The disadvantage, in addition to that of cost,
is that generalisation of the results is difficult. This particular problem is also difficult to solve
with computational methods because of the physical complexity of the problem and the
requirement for validation.
33
What can we say about the external hazard zone?
i) An external hazard zone will only exist if the release within the building is sufficiently large
to raise concentrations within the building above the LFL. For a building of size ~ 1000
m
3
this implies pressurised releases have to exceed ~ 0.1 kg/s and assumes that the
releases take place within the body of the building.
ii) The efflux from the building will comprise a flow of volume oV ~ 3.33 m
3
/s for a building of
size ~ 1000 m
3
and with adequate ventilation and take place through the normal
ventilation openings in the building fabric.
iii) The efflux is a relatively small flow and will be mixed into the recirculation zone behind the
building in the presence of wind. This mixing process is a result of highly unsteady flow
and is wind speed and wind direction dependent but is highly effective.
iv) Hazardous areas will be restricted to the immediate vicinity of vent openings and the
building fabric.
If releases are significantly larger than 0.1 kg/s then a greater hazard will pertain.
For slightly higher flow rates then, if the efflux buoyancy (positive or negative) is not
negligibly small then it will influence the dynamics of the ventilation flow. Specifically it will
preferentially direct hazards to roof or floor and enhance ventilation and if the buoyancy
driven outflow exceeds the ventilation flow a counter flow (additional ventilation) will be
induced.
For much larger flow rates then specific cases need to be considered. Containment and
building effects will keep external concentrations low but events that create an opening in
the fabric of the building, such as opening a door, that make greater ventilation possible
would lead to the outflow of flammable material presenting a localised hazard.
Recommendations:
Enclosure of facilities handling flammable materials should be subject to a hazard
assessment to determine the level of adequate ventilation.
The adequate ventilation recommended in the existing guidance is appropriate to
pressurised releases of Cat A, G(i) fluid of up to ~ 0.1 kg/s for a 1000 m
3
building provided
that the reaction time to respond to a leak is less than ~ 700s.
For an adequately ventilated building the hazard zone is confined to the immediate building
fabric. Escaping vapour will be rapidly entrained and mixed into the recirculation zone
behind the building. This is a region of unsteady but average reverse flow toward the
building at ground level and can be considered to extend to a dimension given by the
smaller of the building height or width. Conventional dispersion models are only accurate at
about 3 -5 times this distance for neutrally buoyant releases.
If ventilation is not adequate then hazard zones will extend at ground level or roof level
depending on the buoyancy of the release. Buoyancy driven flow around vents/openings will
probably enhance the ventilation flow.
34
If ventilation is far from adequate then changing the open area of the building may result in a
large volume of potentially flammable material.
The present guidance is conservative as far as shape factors are concerned but should
relate the size of release and the volume of the building in some more explicit way.
Notes:
The case where a release takes place in a doorway arose in discussion and, by implication,
exists in the present guidelines. The ventilation rate for structures with open doorways will
probably be higher than the 12 changes per hour considered as adequate ventilation.
7. Discussion and Conclusions
In this work we have used tools, developed to investigate the consequence of major
releases, to quantify appropriate guidance for the Area Classification of Petroleum
Installations. The quantification of hazard necessarily starts with specifying the type of
material and the size of release which is very much unknown in the case of small spills and
leaks. Material types have been simulated using 5 example fluid compositions coded (A, B,
C, G(i) and G(ii) following earlier work to update IP 15 and published as an addendum a
Risk Based Approach to Area Classification. That work utilised some very large flow rates
that certainly exceeded any definition of normal spillage. The release rate values used here
necessarily represent the lower end of the hazardous release scale. These should be
larger than arise in normal handling and certainly should not be taken as indicative of the
magnitude of acceptable spills. In all circumstances the potential for spills to occur should
be rigorously assessed and a full hazard assessment carried out where necessary.
It has generally been possible to defend the key recommendations of IP 15 as conservative.
Where revisions are recommended these are strongly dependent on scenario and fluid type.
The major findings of the study are that:
- The hazard radii for pressurised releases of category B and C fluids should be derived
assuming a mechanically generated flammable mist; previously gaseous releases were
assumed. The Hazard radii for category B and C fluids are increased relative to previous
guidance.
- Numeric flammability limits published in Annex D of A Risk-Based Approach to
Hazardous Area Classification for the category B and C fluids have been updated to take
account of the composition of the flammable mist; previously low vapour components
were assumed to rain-out and not contribute to the lower flammability limit evaluation.
- Shape factors for pressurised releases are revised to take better account of the role of
initial jet momentum on the jet trajectory. In particular the lighter than air gases (category
G(i) and G(ii) fluids) are found to have qualitatively more similar shape factors to the two-
phase category A, B and C releases; previously buoyancy was assumed to dominate their
dispersion.
- Hazard radii for discharges from vents are evaluated. The hazard radius varies from
slightly smaller to slightly larger than that in the existing guidance depending on the
properties of the vented vapour.
- The composition of vapour from vents on storage facilities maintained at atmospheric
pressure may be variable (in composition, density and flammability) and the user of the
new guidance should be aware of the effect of this variability because of the
consequence for hazard radii.
35
- The example range of venting rate and vent sizes used in the guidance are not wholly
consistent with the assumption that the discharge takes place at atmospheric pressure.
The relationship between venting rate and pressure of discharge is investigated and a
value of 300 mb suggested as a threshold above which the consequences of pressure
should be assessed. This is of significance for multicomponent fluids where
condensation may occur.
- The existing guidance for liquid spillages is conservative, judged by the volatility of the
model category C fluid. The guidance is applicable to materials with approximately twice
the vapour generation rate of category C fluid under the specimen conditions. Relative
vapour pressures for some common hydrocarbon compounds are listed.
- The existing guidance for sumps is conservative, judged by the volatility of the model
category C fluid.
- Vapour generation at the source of spillage of Category C fluids is a potential hazard
dictated by the spill rate and conditions and not the rate of evaporation of the liquid pool.
The new guidance should emphasise the role of release conditions in determining the
initial vapour generation from spills of category C fluids.
- For releases into confined areas the relative size of spillage and building are of key
importance. The classification Adequate Ventilation has been assessed with respect to
these parameters.
The major uncertainties remaining are:
- The role of liquid rain-out in influencing flammability hazards for releases from low (~ 5
bar) processes.
- How properly to advise the effect of different fluid compositions in hazards.
- How to properly balance guidance for area classification and necessary procedures for
hazard assessment.
8. References
A Risk-Based Approach to Hazardous Area Classification, Institute of Petroleum, November
1998.
Shell FRED 3.1 User and Technical Manual, Shell Global Solutions, (2000).
HGSYSTEM v3.0; Technical Reference. Shell Global Solutions (1996) available in
electronic form (pdf) at http://www.hgsystem.com/.
Johnson D.W. and Woodward J.L.; Release: a model with data to predict Aerosol Rainout in
Accidental Releases. AIChE. (1999).
Ramsdale S., Tickle G. Review of Release rain-out model and the Center for Chemical
Process Safety (CCPS) data, Contract Research report 277/2000, Publ. HSE Books.
Lees F. Loss prevention in the Process Industries, ed 2. Publ. Butterworth/Heinemann,
1996.
Bearman P.W., 1972, "Some Recent Measurements of the Flow Around Bluff Bodies in
Smooth and Turbulent Streams", Paper presented at a synopsium on external flows,
University of Bristol.
36
Britter, R.E. Hunt, J.C.R. and Puttock, J.S., 1976, "Predicting Pollution Concentrations Near
Buildings and Hills." Presented at the Institution of Measurements and Control Conference
on Systems and Models in Air and Water Pollution, London, Sep 22-24, pp 7-1 to 7-15.
Castro, I.P., & Robins, A.G., 1977, "The Flow Around a Surface-Mounted Cube in Uniform
and Turbulent Streams", J. Fluid Mech., vol. 79, part 2, pp. 307-335.
Hosker, R.P., 1979, "Empirical Estimation of Wake Cavity Size Behind Block Type
Structures." In the reprints of Fourth Symposium on Turbulence, Diffusion and Air Pollution,
Reno, NV, Jan. 15-18, pp 603-609.
Snyder, W.H. and Lawson, R.E., Jr., 1994, "Wind tunnel measurements of flow fields in the
vicinity of buildings", Reprint vol.: 8th AMS Conf. on Appl. Air Poll. Meteor., with AWMA, Jan.
23-28, Nashville, TN.
37
Appendix A: Methodology
A1. Calculating Flammability limits.
The flammability limit of a mixture, L (%v), should therefore be computed by first using Le
Chateliers law to establish the volumetric flammability limit of a vapour only mixture:
1
L
y
L
i
i
=
_
Where y
i
denotes the mol fraction of the species i and L
i
the species vapour phase
flammability limit. This flammability limit can then be converted to a mass-based criteria -
most simply by the assumption of an ideal gas whereby the mixture density is given by:
mix i i
y M
P
RT
=
_
kg/m
3
where M
i
are the species molecular weights, P the pressure, R the ideal gas constant and T
the temperature.
The mass based limit is.
L
L
m mix
=
100
kg/m
3
The calculation procedure for flammability limits is embodied in the Shell FRED 3.0 (and
higher) software tool available from Shell Global Solutions.
A2. Calculating hazard radii for pressurised releases.
Hazard radii need to be calculated using a multicomponent and multi-phase dispersion
model. Shell Global Solutions has developed such a capability as part of HGSYSTEM which
is a set of dispersion models, developed at the specification of industry consortia, and made
publicly available via the internet. HGSYSTEM has been used as a reference model in
model evaluation exercises and the model components set a standard against which
commercial models are tested. Model vendors should be able to relate the performance and
physical basis of their models to HGSYSTEM.
The calculations used the fluid properties given in Table 2 of the main report and the
following conditions:
Storage Temperature - 20 C
Ambient Temperature - 30 C
Wind Speed - 2 m/s
Stability Class - D
Roughness - 0.03 m
Horizontal Release @ 5 m for R
1
Horizontal Release @ 1 m for R
2
38
Of these the surface roughness length 0.03 m has been increased from the value of 0.003 m
used in the original guidance. The 2-phase jet calculations used for the hazard radii are not
sensitive to this parameter so the change is semantic rather than critical. A value of 0.003 m
is appropriate to extremely flat terrain with even and low lying vegetation. This is not
characteristic of process sites where values as large as 0.3 m could in some circumstances
be justified. The choice of 0.03 m is at the lower end of realistic values for process sites and
is conservative. The role of roughness length in dispersion models is to characterise the
ambient turbulence. This is the dominant mixing mechanism for low-momentum releases.
For convenience, and following the methodology in the previous update to IP 15, the
discharge rates were calculated using the Shell Global Solutions Generalised Release
Model (GENREL) incorporated in FRED 3.1 and setting pipe friction to be negligible. The
HGSYSTEM model AEROPLUME has a simpler release model and, for two phase releases,
generates a range of possible mass flow rates for a set of input conditions rather than a
single value. The GENREL and AEROPLUME suggested rates are, for most mixtures,
comparable.
The hazard radii were calculated using the two-phase jet dispersion model AEROPLUME.
AEROPLUME calculates cross-sectional average concentrations. The hazard radius R
1
should be taken as the distance at which the average concentration has fallen to a value of
0.7 LFL. The hazard radius R
1
should be taken as the distance at which the average
concentration has fallen to a value of 0.5 LFL. This arises because the hazard radius is
evaluated with respect to the maximum concentration in the jet and the relationship between
the maximum and average jet concentrations depends upon the shape of the concentration
profile.
Within FRED 3.1 the AEROPLUME results are automatically processed to take account of
the concentration profiles and give hazard radii directly and in a manner transparent to the
user. We note that a key result of this study, that a mass-based flammability limit should be
used for two-phase flow calculations; is not implemented in the post-processor tools
supplied with HGSYSTEM 3.0. The post-processor PROFILE from v 3.0 should not be
used for high flash-point fluids.
A3. Calculating Hazard Radii for vents.
Ambient conditions for calculating the hazard radii for vents were the same as for the
pressurised releases. The key feature of the vent flows is that they are assumed to be ideal
gases, i.e. no account is taken of the potential for two phase flow.
The dispersion calculations were carried out using AEROPLUME assuming a vertical
release. The hazard radius was defined as:
( )
( ) R Z H X = +
2 2
where H was the assumed vent height of 10 m and Z, X the vertical and downstream co-
ordinates of the point of maximum concentration in the jet equal to the flammability limit.
This is the centroid position of the jet calculated by AEROPLUME with an average
concentration equal to 0.7 LFL.
39
A4. Calculating pool evaporation and dispersion
The pool evaporation calculations were carried out using the liquid spills model LPOOL
which is part of HGSYSTEM. LPOOL is an implementation of the ExxonMobil Liquid Spills
Model LSM90. It takes account of the multicomponent evaporation of material either spilled
into a bund or onto an unconstrained surface.
The basis of pool evaporation models is set out in Lees (1996). The evaporation rate is
based on a forced convection analogy and increases with wind speed and with pool area.
For non-boiling pools the evaporation rate increases with the vapour pressure of the spilled
liquid. A pool model adds a liquid spread, a heat balance and a component balance to the
base evaporation correlation.
For consistency with the other hazard radii a wind speed of 2 m/s wind speed and an
ambient temperature of 30 C was used. To enhance the evaporation rate further,
maximum insolation (spill at mid-day, cloudless sky) was assumed. The insolation factors in
LPOOL are conservative (high) for mid-latitudes.
As noted in the text, pool evaporation is a transient process that gives rise to a time-varying
gas cloud. In particular we found that the highest evaporation rates came during the first
moments of the spill when the lightest components were evaporating.
Calculations made using the HGSYSTEM unsteady dense gas dispersion model HEGADAS-
T showed that the vapour cloud dispersed extremely quickly and, with the target conditions,
it was not possible to attain the hazard radii in the previous guidance.
We then carried out a sensitivity test with the HGSYSTEM steady state dense gas
dispersion model HEGADAS-S to determine what continuous release rates of dense gas
would be consistent with the existing guidance. We found, as given in the main text, that
these rates are higher than the vapour rates for category C spills.
40
Appendix B:
Preliminary Investigation of Hazard Radii and Shape Factors
for the revision of IP15: the Area Classification Code for
Petroleum Installations.
Summary
As part of the revision of the Institute of Petroleum publication Area Classification Code for
Petroleum Installations, more commonly known as IP 15, dispersion calculations were
carried out to cross check tables D3.2 and D3.3 of the publication A risk-based Approach
to Hazardous Area Classification.
The objectives of revisiting the calculations were twofold:
- To provide a subset of results more in line with petroleum installation needs.
- To verify shape factors for releases - especially those representing a refinery hydrogen
stream.
In repeating the calculations it was noted that the usual practice of specifying flammability
limits in terms of volumetric concentrations for gases leads to possibly incorrect and too low
estimates of hazard distances for class B and class C liquids. These are materials that have
high flash points at ambient conditions but may be placed under high pressure when
pumped from one location to another. The leak of high pressure liquid may give rise to a
flammable mist or, in any event, lead to the jetting of potentially flammable liquid over
distances much greater than those given in the present IP 15 guidance.
This note describes the revised calculations and discusses the problem of how to assess the
hazard range for high flash-point fluids. The calculations in this report were carried out with
the HGSYSTEM models DATAPROP, AEROPLUME and PROFILE to facilitate the
presentation of the many results. These models are equivalent to the components of the
FRED model version 2.3.
As a result of this work the definition of flammable limit used in deriving contours with the
software package FRED has been changed. FRED versions 3.0 and higher are compatible
with these results.
B1. Introduction
Hazardous area classification requires that an assessment be made of the extent of the
zone enclosing an operational area where flammable materials are handled and where small
leaks might occur unnoticed for a short period of time. With this knowledge precautionary
steps can be taken to reduce the potential for ignition, for example by placing constraints on
electrical installations.
The Institute of Petroleum (IP) publication: A risk-based Approach to Hazardous Area
Classification adopted (in Annex D) tables of hazardous distances from an internal Shell
source, which contained calculations made with the FRED dispersion model, version 2.1,
that were appropriate to an exploration and production environment where high pressures
and hence high flow rates are the norm. These tables included events that would only arise
as a result of mechanical damage on a petroleum refinery and which would in any case
demand an emergency response. Including these events as part of advice on Area
41
Classification gives a misleading impression of the hazard distance to be associated with
normal operation.
This work was intended to rework the content of current IP publication advice and verify that
the shape factors (Figure 6.2 and Figure 6.3 of IP 15) for dense and buoyant gas releases
were realistic. It was expected that the advice would be essentially unchanged. However, It
was noticed that the published data gave very small (more than ten times smaller) hazard
distances for the Class C
vi
fluid, compared with the Class A fluid; for example:
Release Conditions Hazard radius (m)
Pressure (bar) Diameter (mm) category A category C
100 10 39 3
Table B.1 Extract from Annex D of A Risk Based Approach to Hazardous Area
Classification
Physical intuition suggests that, to take the example above, a the hazard radius of 3 m is far
too small compared with the likely throw of a non-volatile but flammable liquid jet driven by
a 100 bar pressure through a 10 mm hole. There have been discussions and project
proposals to investigate the throw of water jets for area clarification and electrical safety
purposes and unpublished work carried out for ICI has shown that small holes and low drive
pressures can effectively transport material well beyond the present category C fluid
guideline distances.
Pressurised releases can also atomise material with a high flash-point to give rise to mists
that are flammable. The flammability limits for fine aerosols are akin to those for vapours
although there is a paucity of informational data in the literature. It is believed that, in
general, the flammability limit of fine aerosols, when expressed as the mass of fuel per unit
volume, is similar to the flammability of the vapour alone (Lees, F.J. Loss Prevention in the
Process Industries, 2
nd
ed., vol. 2, section 16.4.3). This is significant because it is common
practice is to use volumetric units to express flammability limits for gases and for mixtures of
gases. For a given mass concentration the equivalent volume fraction is substantially
reduced in the presence of a small liquid component because of the very large density
difference between the two phases. The use of a volumetric criterion for evaluating the
hazard distance may therefore be misleading when two-phase mixtures are involved and,
worse, give too small an estimate of the hazard distance.
In this note we compare the use of mass and volume based flammability criteria for
establishing hazard zones on the assumption that releases of high flashpoint material will
form a flammable mist and hence that their hazard can be addressed in a comparable
manner to gaseous and low flash-point 2-phase releases. The issue remains of whether
more coherent jets exist and are able to deliver flammable liquid to an even more distant
target but we do not attempt to quantify this here.
We should also comment upon the mitigating effect of liquid droplet rainout from jet
releases. For high flashpoint liquids that are atomised and ejected as jets the process of
droplet collision will lead to a growth in the droplet size with time and the subsequent
deposition of liquid. An extensive body of work has been carried out for single component
fluids by the American Institute of Chemical Engineers (AIChE) and a release model
developed to predict the loss of liquid from such events (Johnson, D.W., Woodward, J.L.,
Release: A model with data to predict Aerosol Rainout in Accidental Releases, AIChE,
1999). For area classification purposes we are considering material transport over very
short distances and with transit times of a few seconds and over trajectories that are
vi
Model fluid compositions are given in Table B2,
42
determined almost entirely by the initial jet momentum and orientation. To give conservative
estimates of the hazard zone it is reasonable to assume that liquid droplet rainout in the
vicinity of the source is a second order effect. For the larger releases that appear in risk
assessment scenarios where fluid transport times of several minutes may occur, the AIChE
work should be reviewed to see if accounting for droplet rainout makes a substantial
difference to model predictions.
B2. Calculation Procedure.
Of necessity only a limited number of variables can be considered in this investigation. The
area classification process is NOT a full hazard assessment which takes account of the
range of meteorological and process conditions, release events etc. Accordingly a single set
of release conditions was chosen and these are the same as those used in the IP addendum
A risk-based Approach to Hazardous Area Classification with the exception that the
surface roughness length has been increased by a factor of 10 from an unreasonable
overland value to a value that is still low (by a factor of 10) compared to that used for
process sites. This has no significant effect on any of the calculations but makes for a more
physically realistic input data set.
The fluid compositions used are given in Table B2 and the base calculation conditions in
Table B3. The two parameters in Table B3 that were varied to investigate the shape of the
hazardous zone were the release orientation and the release height.
A reduced set of operating conditions were assumed. Four drive pressures (100, 50, 10, 5)
bara. and five hole sizes (1, 2, 5, 10 , 20) mm (as diameters). The shape factors for the
releases were calculated using the largest hole size and the greatest drive pressure.
43
Stream
Component
(mol perc)
Cat. A Cat. B Cat. C Cat. G
(I)
Cat. G
(ii)
Comp.
LFL
MW Boiling
point
o
C
N
2
Nitrogen 0 0 0 2 2 - 28.01 -196
C
1
Methane 0 4 0 88.45 10 5.3 16.04 -161
C
2
Ethane 0 0 0 4.5 3 3 30.07 -87
C
3
Propane 70 6 1 3 3 2.2 44.09 -42
C
4
Butane 30 7 1 1 1 1.86 58.12 -1
C
5
Ethane 0 9 2 1 0 1.50 72.15 36
C
6
Hexane 0 11 3 0 0 1.2 86.17 69
C
7
Heptane 0 16 3 0 0 1.2 100.20 98
C
8
Octane 0 22 27 0 0 0.95 114.23 126
C
9
Nonane 0 0 25 0 0 0.83 128.26 151
C
10
Decane 0 25 38 0 0 0.77 142.28 173
H
2
O Water 0 0 0 0.05 0 - 18.02 100
CO
2
0 0 0 0 1 - 44.01 -78(sub)
Hydrogen 0 0 0 0 80 4.1 2.02 -253
Average
MW
48.3 100.06 125.03 18.74 7.03
LFL (vol%) 2.09 1.70 1.52 4.82 4.03
Table B.2 - Definition of category Fluids and their properties. The vapour phase
mixture flammability limits are taken from the IP addendum publication and were
evaluated using Le Chateliers Law.
Standard conditions Base Case values Range of values
Ambient temperature 20
o
C
Relative humidity 70 %
Wind speed 2 m/s
Reference Height 10.0 m
Stability class D
Surface roughness 0.03 m
Sample time 18.75 s (~ instantaneous)
Release height 1.0 m 0 to 15 m
Reservoir Temperature 20
o
C
Release angle (relative to horizontal) 0
o
-30
o
to 90
o
Table B.3 - Base parameters for the calculation
Release rates for the five test fluids were evaluated using the Generalised Release Model
from FRED 2.3 and assuming negligible friction losses from any pipe runs. The release
rates calculated in this way have proved more consistent than the default release rate used
by the Aeroplume model which is a literature correlation for two phase flow. Aeroplume also
has a built in two phase discharge mode, used for advice purposes under normal program
execution, and this usually agrees to within a few percent of the Generalised Release Model
predictions.
As expected it was found that the release mass flux per unit area was hole size independent
for the release conditions and the release rates are plotted as a function of hole size and
pressure for each of the fluids in Figure B2. The results are straightforward except for
category A fluid at the lowest drive pressure. Here the standardised 5 bar drive pressure at
ambient temperature is below the saturated vapour pressure of propane and above the
44
bubble point of propane
vii
. This means that in a vessel the category A vessel would exist as
discrete two phases giving three discharge scenarios: a vapour only release of a propane
rich vapour; a liquid only release of a butane rich butane/propane mix and a two phase
release; all depending on the location of the release point. Any significant length of piping
between the storage vessel and the release would favour a two phase release because
dissolved propane would vaporise in the pipe as the pressure decreases. We have
therefore carried out two sets of calculations for this low pressure scenario. A high mass
rate discharge of a 45:55 propane/butane mixture corresponding to the liquid phase
composition of the 70:30 propane/butane total mixture and a low mass rate discharge of the
original 70:30 propane/butane mixture as a 2 phase release. Figure B1 shows mass flow
rates for both circumstances.
0.1
1
10
0 2 4 6 8 10 12 14 16 18 20
R
e
l
e
a
s
e
R
a
t
e
k
g
/
s
Hole size (dia.), mm
Release Rate as a function of hole size and drive pressure
Category A Fluid
100 bar
50 bar
10 bar
5 bar - liquid
5 bar - 2 phase
Figure B.1: Release rate as a function of hole size and drive pressure for each of the five
conditions, category A fluid.
vii
In the Addendum to IP 15 it was recognised but not explicitly stated that the 5 bar scenario was
ambiguous. The approach there was to decrease the ambient temperature for this run and hence the
saturation vapour pressure of propane.
45
0.1
1
10
0 2 4 6 8 10 12 14 16 18 20
R
e
l
e
a
s
e
R
a
t
e
k
g
/
s
Hole size (dia.), mm
Release Rate as a function of hole size and drive pressure
Category B Fluid
100 bar
50 bar
10 bar
5 bar
0.1
1
10
0 2 4 6 8 10 12 14 16 18 20
R
e
l
e
a
s
e
R
a
t
e
k
g
/
s
Hole size (dia.), mm
Release Rate as a function of hole size and drive pressure
Category C Fluid
100 bar
50 bar
10 bar
5 bar
Figure B.2: Release rate as a function of hole size and drive pressure for each of the conditions,
categories B and C fluids.
46
0.1
1
10
0 2 4 6 8 10 12 14 16 18 20
R
e
l
e
a
s
e
R
a
t
e
k
g
/
s
Hole size (dia.), mm
Release Rate as a function of hole size and drive pressure
Category G(i) Fluid
100 bar
50 bar
10 bar
5 bar
0.1
1
10
0 2 4 6 8 10 12 14 16 18 20
R
e
l
e
a
s
e
R
a
t
e
k
g
/
s
Hole size (dia.), mm
Release Rate as a function of hole size and drive pressure
Category G(ii) Fluid
100 bar
50 bar
10 bar
5 bar
Figure B.3 - Release rate as a function of hole size and drive pressure for each of the conditions,
categories G(i) and G(ii) fluids.
47
The AEROPLUME model was run using automatically generated input files for the different
operating conditions. AEROPLUME outputs a cross-sectional average jet concentration. To
derive concentration contours the output of the AEROPLUME model can be post-processed
using the PROFILE program. This superposes a Gaussian profile upon the AEROPLUME
results and constructs the required contour. This is the normal operating procedure in the
FRED model where the results are automatically displayed as a graphic and the underlying
calculations are not visible to the user.
It is important to note that PROFILE assumes a ground-reflected Gaussian distribution which
means that the relationship between the maximum plume concentration and the average
plume concentration depends upon the height of the plume above ground. Where we are
just interested in the furthest extent of the maximum LFL concentration from the source we
can estimate this directly from AEROPLUME by calculating the distance to a slightly smaller
concentration. The ratios are shown in Table B.4.
Peak Concentration Averaged Concentration
Elevated Plume LFL 0.7 LFL
Grounded Plume LFL 0.5 LFL
Table B.4 : Equivalence between peak and averaged concentrations in the HGSYSTEM and
FRED models.
B3. Effect of using mass and concentration based LFL criteria:
The mass equivalent LFL was derived from the vapour properties in Table B.2 by multiplying
the vapour phase volume fraction by the vapour phase material density assuming an ideal
gas, one atmosphere pressure and the ambient temperature.
The downwind LFL position on a mass and on a volume basis was determined by linear
interpolation of distance between bounding entries in the AEROPLUME output file and using
the ratios given in Table B.4. This is adequate for the purposes of this study.
Figure B.4 shows a composite of all results for a 1 m horizontal release. The distance to
LFL calculated using the mass concentration and the distance to LFL using the volume
fraction criteria are plotted against each other.
- For the two gas mixtures (category G(i) and G(ii) ) the flammability limits equivalent and
an identity results.
In the presence of a liquid phase, volumetric concentrations underestimate the amount of
flammable material present because the contribution of liquid to volume is negligible. We
see:
- For the low flashpoint material (fluid A) sufficient liquid has evaporated for the jet to be
wholly gaseous at LFL so the two flammability definitions and calculated LFL distances
are equivalent
- For fluid category B the volume based definition noticeably understates the hazard
distance by ~ 30%.
- For fluid category C there is almost an order of magnitude difference in the hazard
radius.
48
0.1
1
10
100
0.1 1 10 100
D
i
s
t
a
n
c
e
t
o
L
F
L
u
s
i
n
g
v
o
l
u
m
e
f
r
a
c
t
i
o
n
Distance to LFL using mass concentration
Influence of fluid type on LFL definition
A
B
C
Gi
Gii
Figure B.4 : Effect of using mass concentration and volume concentration definitions on
distance to LFL for the five fluids.
This is a significant finding. It suggests that using volume based flammability criteria to
derive distances to LFL may lead to underestimating the hazard of two-phase releases.
To illustrate this more clearly on a linear scale. Figure B.5 shows some example shape
factor calculations for the category B fluid. An elevated release (15 m height) is assumed
and calculations carried out for different release directions. The resulting overlapping cigar-
shaped contours are results that would be obtained by following normal practice and using
volumetric limits (e.g. with the FRED 2.3 model). The points show where the maximum
distance to LFL occurs using the mass based criterion.
49
0
10
20
30
40
50
0 10 20 30 40 50
H
e
i
g
h
t
,
m
Downwind distance, m
IsoConcentration contours for CATB fluid, 100 bara, 20 mm hole
Releases at different inclinations
Values from Aeroplume
Figure B.5 - Visual Example of the difference in using volumetric (contours) and mass
concentration definitions (points) for the LFL of a category B fluid.
B.4 Shape factors
Shape factors for the hazardous area are based on calculations of jet dispersion in different
directions and the existing classification advice implicitly assumes that, because of the
effects of body forces, the shape factor for buoyant releases is different to that for dense gas
releases.
This work shows that the hazard radius for the different fluids is determined by jet
momentum rather than body forces, but that there is an effect of the ground on entrainment
which is different for dense and for buoyant releases.
Figure B.6 shows the shape factor for a category A fluid released from a 1 m height
(maximum flow rates are used in all of these calculations). It is assumed that the wind
direction is from the left of the picture. For a category A fluid there is no effect of liquid load
on the calculation procedure and we have included both the full iso-concentration contours
as would be obtained using the FRED model and the points derived from the AEROPLUME
model using the mass concentration definition of LFL.
For releases that do not touch the ground the hazard range is essentially a flattened circular
locus. An increasing wind speed would flatten the shape further by decreasing the vertical
penetration of the jet because the contribution to entrainment from cross-wind mixing is
increased.
50
For releases that do touch the ground the dispersion distance is increased on two counts:
- Within the assumptions of the jet model, the overall entrainment is reduced. Initially the
perimeter area of the jet is reduced by the physical presence of the ground and then, as it
spreads out over the ground and slows down, the entrainment is inhibited by the density
difference between the plume and the air.
- The reflected Gaussian shape of the concentration distribution means that the ratio of the
peak concentration to the average concentration within the jet changes when the jet hits
the ground. For an elevated plume the peak concentration is ~1.42 times greater than
the average concentration across the plume and occurs on the plume centroid. For a
grounded plume the peak concentration is ~ 2 times the averaged concentration and
occurs at the ground. The locus of points from AEROPLUME shown in Figure B.6 reflect
this assumption.
0
10
20
30
40
50
0 20 40 60 80 100
H
e
i
g
h
t
,
m
Downwind distance, m
IsoConcentration contours for CATA fluid, 100 bara, 20 mm hole
Releases at different inclinations
Centroid Values from Aeroplume
Peak Values from Aeroplume
Figure B.6 : LFL isopleths for a category A fluid released at angles from -20 to 90 degrees to
the horizontal and centroid position of LFL from Aeroplume (points). The ground level LFL
concentration position from Aeroplume for the three lowest trajectory releases are shown as
vertical bars.
These two factors substantially increase the hazard distance for horizontal releases near to
the ground. This methodology is conservative because extra mixing caused by the friction
between the jet and the ground is neglected. Also it results in a step change increase in
hazard radius for releases that only just touch the ground as shown in Figure B.7 below.
51
0
10
20
30
40
50
0 20 40 60 80 100
H
e
i
g
h
t
,
m
Downwind distance, m
IsoConcentration contours for CATA fluid, 100 bara, 20 mm hole
Releases at different inclinations
Values from Aeroplume
Peak Values from Aeroplume
Figure B.7 : Shape factor for an elevated release of category A fluid
Figure B.7 shows the effect of increasing the release height to 15 m. Calculations to the
required concentrations can only be made for release angles greater than -20 degrees with
the present version of the AEROPLUME model which does not model steep impacts of high
momentum jets. We do not expect these steeper impacts to lead to greater dispersion
distances as the impact will lead to the generation of additional turbulence which will
increase the mixing rate.
Figure B.8 shows the shape factor for the release of category G(i) fluid from a height of 1 m
We find that the distance to LFL along the plume centroid is nearly independent of release
angle.
The distance to LFL at the ground for those releases that hit the ground is again greater than
that of the elevated jet so that the overall shape factor is similar to that of a dense gas
release. The effect of the ground is less dramatic because the spreading fluid is less dense
than the air and is convectively unstable. This means that entrainment into the grounded
lighter than air gas is greater than it would be for a dense gas. Consequently the hazard
distance at the ground is not as great.
52
0
5
10
15
20
0 5 10 15 20
Height, m
Downwind distance, m
IsoConcentration contours for CATG(i) fluid, 100 bara, 20 mm
Releases at 90, 70, 50, 30, 10, 0,-10 degrees from the horizontal
Centroid Values from Aeroplume
Ground level Values from Aeroplume
Figure B.8 : Shape factor for the category G(i) Fluid. Points are centroid distance to LFL and
the vertical bars show the ground level distance to LFL for the two lowest trajectory releases.
The category G(ii) fluid is markedly more buoyant than the category G(i) fluid and to
investigate whether the extra buoyancy significantly modifies the shape factor these have
been derived for release heights of 1m, 5 m, 10 m and 15 m.
Figure B.9 shows that, for a 1 m release height the results are qualitatively very similar to
those for a category G(i) fluid. The grounded jets entrain air more efficiently because the
flow is more unstable and there is less enhancement of the distance to LFL at the ground..
Increasing the release height to 5 m barely alters the shape factor. As the release height is
increased to 10 m the basic shape is still maintained with a downward directed jet at -30
degrees hitting the ground before diluting to LFL. The ground footprint is very similar to that
for the 1 m release. Increasing the release height to 15 m again results in a fan shaped
hazard range that is almost symmetric about the release height showing that, in the absence
of ground contact, the hazard radius is determined by the jet orientation. This contrasts with
Figure 6.3 of the IP 15 publication that suggests that there is a smaller hazard zone below
the release point than above it.
53
0
5
10
15
20
25
30
35
0 5 10 15 20 25 30 35
Height, m
Downwind distance, m
LFL Concentration contours for CATG(ii) fluid
100 bara, 20 mm hole
Centroid LFL position from Aeroplume
Groundlevel LFL from Aeroplume
0
5
10
15
20
25
30
35
0 5 10 15 20 25 30 35
Height, m
Downwind distance, m
LFL Concentration contours for CATG(ii) fluid
100 bara, 20 mm hole
Centroid LFL position from Aeroplume
Groundlevel LFL from Aeroplume
Release height: 1 m
Release height: 5 m
54
Figure B.9 : Shape Factors for a release of Cat. G(ii) fluid released from a height of 1 m, 5m , 10 m and 15
m above the ground. Points indicate the plume centroid position at LFL and the vertical bars the ground level
distance to LFL for the three lowest trajectories all derived from the Aeroplume model. The isopleths are
calculated with the Profile model and are equivalent to the output from the FRED 2.3 model.
0
5
10
15
20
25
30
35
0 5 10 15 20 25 30 35
H
e
i
g
h
t
,
m
Downwind distance, m
LFL Concentration contours for CATG(ii) fluid
100 bara, 20 mm hole
Centroid LFL position from Aeroplume
Groundlevel LFL from Aeroplume
Release height: 10 m
0
5
10
15
20
25
30
35
0 5 10 15 20 25 30 35
H
e
i
g
h
t
,
m
Downwind distance, m
LFL Concentration contours for CATG(ii) fluid
100 bara, 20 mm hole
Centroid LFL position from Aeroplume
Release height: 15 m
55
B.5 Hazard Radii
The calculations carried out for each of the releases suggest that for elevated jets the shape
factor should comprise a semi-spherical hazard zone evaluated as the distance to LFL on
the plume centroid for a horizontal jet in the downwind direction. The calculations given in
Tables B.5 & B.6 show that the effect of the ground is significant for low level releases of all
materials and therefore the shape factor should be extended at the ground.
Distance to LFL at the plume centroid height for a 1 m high horizontal release, m
Fluid Category Pressure
(bara)
20 mm 10 mm 5mm 2 mm 1mm
A 100 59.6 27.8 11.1 4.0 2.0
50 59.8 28.0 10.9 3.9 2.0
10 60.2 28.4 10.8 3.4 1.8
5 63.5 29.2 10.9 3.1 1.7
B 100 39.2 17.4 7.0 2.9 1.5
50 39.0 17.2 6.7 2.8 1.4
10 30.0 12.5 5.0 2.1 1.1
5 23.4 8.6 3.9 1.7 0.9
C 100 36.6 16.0 6.6 2.7 1.4
50 36.7 16.0 6.4 2.7 1.4
10 37.0 16.4 5.8 2.5 1.3
5 37.2 16.6 5.3 2.4 1.3
G(i) 100 13.2 6.4 3.3 1.3 0.7
50 8.4 4.2 2.1 0.9 0.4
10 3.5 1.8 0.9 0.4 0.2
5 2.6 1.3 0.7 0.3 0.2
G(ii) 100 15.6 9.3 4.9 2.0 1.0
50 12.3 6.8 3.5 1.4 0.7
10 6.3 3.3 1.6 0.7 0.3
5 4.7 2.4 1.2 0.5 0.3
Table B.5 : Summary of Distances to LFL at the plume centroid height for the horizontal
release from a height of 1 m.
56
Distance to the LFL at Ground Level for a 1 m high horizontal release, m.
Fluid Category Pressure
(bara)
20 mm 10 mm 5mm 2 mm 1mm
A 100 83.0 39.8 17.1 5.1 2.7
50 83.1 40.2 17.1 4.9 2.6
10 85.5 41.0 17.2 4.1 2.3
5 93.5 42.7 17.3 3.7 2.1
B 100 54.6 25.2 9.9 3.8 1.9
50 54.2 25.0 9.6 3.6 1.9
10 42.0 18.8 6.3 2.7 1.4
5 33.1 13.8 4.8 2.1 1.1
C 100 50.7 23.2 9.1 3.5 1.8
50 50.9 23.4 8.8 3.4 1.8
10 51.0 23.8 8.4 3.1 1.6
5 52.3 24.0 8.3 2.9 1.6
G(i) 100 18.7 9.2 4.4 1.8 0.9
50 12.0 5.7 2.9 1.2 0.6
10 4.9 2.5 1.2 0.5 0.3
5 3.6 1.8 0.9 0.4 0.2
G(ii) 100 21.3 12.8 6.6 2.8 1.4
50 16.7 9.3 4.7 2.0 1.0
10 8.6 4.4 2.3 0.9 0.5
5 6.3 3.2 1.7 0.7 0.4
Table B.6 : Distance to LFL at ground level for the horizontal release from a height of 1 m.
B6. Conclusions
The purpose of this study was to verify shape factors and to cross check tables D3.2 and
D3.3. of the publication A risk-based Approach to Hazardous Area Classification with a set
of conditions more in line with petroleum installation needs.
Dispersion calculations were carried out for 5 types of fluids (Categories A, B, C, G(i) and
G(ii)) under a range of pressure and release rate conditions.
We found that:
The usual practice of specifying flammability limits in terms of volumetric concentrations for
gases leads to possibly incorrect and too low estimates of hazard distances for class B and
class C liquids. These are materials that have high flash points at ambient conditions but
may be placed under high pressure when pumped from one location to another. The leak of
high pressure liquid may give rise to a flammable mist or, in any event, lead to the jetting of
potentially flammable liquid over distances much greater than those given in the present IP
15 guidance.
Shape factors for the hazardous area, based on calculations of jet dispersion in different
directions, show that the hazard radius for the different fluids is determined by jet momentum
rather than body forces. Consequently, the shape factor for buoyant releases is qualitatively
similar to that for dense gas releases.
57
For releases that do not touch the ground, the hazard radius is nearly independent of
release angle and the hazard range is essentially a circular locus, slightly flattened at the top
owing to the wind. For a buoyant elevated release, the hazard radius is almost symmetric
about the release height. This contrasts with Figure 6.3 of the IP15 publication that suggests
that there is a smaller hazard zone below the release point than above it.
We recommend that:
Mass based flammability criteria should be used for multiphase releases. The Shell Global
Solutions FRED model (version 3.0 and higher) has been modified on this basis.
The shape factor should comprise a semi-spherical hazard zone evaluated as the distance
to LFL on the plume centroid height for a horizontal jet in the downwind direction. There
should be an extra margin of allowance for grounded jets.