...

RR1028 Further validation of the ACE instantaneous source model

by user

on
Category: Documents
5

views

Report

Comments

Transcript

RR1028 Further validation of the ACE instantaneous source model
Health and Safety
Executive
Further validation of the ACE instantaneous
source model
Prepared by the Health and Safety Laboratory
for the Health and Safety Executive 2014
RR1028
Research Report
Health and Safety
Executive
Further validation of the ACE instantaneous
source model
Simon Coldrick
Health and Safety Laboratory
Harpur Hill
Buxton
Derbyshire SK17 9JN
ACE (Airborne Concentration Estimate) is a model originally developed by WS Atkins on behalf of HSE for
computing source terms arising from instantaneous flashing releases. The model has since been reviewed
and enhanced through a series of extensive tests that have been performed by HSL. Inputs to ACE are the
substance, its mass and storage conditions and ambient weather conditions. The model then computes
the formation of a vapour cloud assuming total loss of containment. This is done in two distinct stages; an
explosion phase, which models the initial flashing process and a turbulent growth stage, describing further
radial expansion and air entrainment. The resulting cloud properties can be used as the input to an atmospheric
dispersion model. This report describes an updated validation of ACE in which overall cloud radius predictions
have been compared with those obtained from experiment. In addition, an attempt has been made to further
the validation by comparing cloud expansion velocities. This was not previously carried out for ACE and was
considered an area where the validation could be enhanced.
This report and the work it describes were funded by the Health and Safety Executive (HSE). Its contents,
including any opinions and/or conclusions expressed, are those of the authors alone and do not necessarily
reflect HSE policy.
HSE Books
© Crown copyright 2014
First published 2014
You may reuse this information (not including logos) free
of charge in any format or medium, under the terms of the
Open Government Licence. To view the licence visit
www.nationalarchives.gov.uk/doc/open-government-licence/,
write to the Information Policy Team, The National Archives, Kew,
London TW9 4DU, or email [email protected]
Some images and illustrations may not be owned by the
Crown so cannot be reproduced without permission of the
copyright owner. Enquiries should be sent to
[email protected]
ii
CONTENTS
1
INTRODUCTION ..................................................................................... 1
2
2.1
2.2
2.3
2.4
VALIDATION ........................................................................................... 2
Original validation .................................................................................... 2
Experimental data .................................................................................... 3
Modification of ACE output ...................................................................... 7
Validation results ..................................................................................... 8
3
CONCLUSIONS .................................................................................... 23
4
REFERENCES ...................................................................................... 24
iii
EXECUTIVE SUMMARY
Until recently, HSE have employed the IRATE model to compute source terms arising from
instantaneous releases of superheated liquids. Following review of the IRATE model, (Tickle,
2012a) its use has been discontinued. In its place, HSE are considering the use of the ACE
(Airborne Concentration Estimate) model developed by WS Atkins (Gilham et al., 1997a,b).
ACE models the formation of a vapour cloud in two distinct stages:
1) an explosion phase, which models the initial flashing process.
2) a turbulent growth stage, describing further radial expansion and air entrainment. Cloud
properties can be obtained at the end of the turbulent growth stage.
Objectives
Shepherd and Deaves (2000) describe a limited validation of ACE but acknowledge that
validation is difficult due to the limited experimental data and the difficulty in its interpretation
in relation to the various stages of the release. Therefore, HSE requested HSL to review the
validation of ACE in light of any new experimental data, or by more in-depth comparison with
existing data.
Main Findings
Comparison of overall cloud radius predictions against experiments at a range of scales showed
that the ACE predictions were in reasonable agreement. Particularly good agreement was
obtained against the large scale experiments of Hardee and Lee (1975) and Bettis and Jagger
(1992). For the other tests, ACE produced much larger cloud radii. The main deficiency of
absolute radius comparisons is that it is not clear in many cases what the final cloud radius was;
either the end of the measuring scale was reached (e.g. Pettitt, 1990) or the cloud was ignited
(Maurer et al., 1977). Using the default turbulent velocity value of 10 m/s resulted in slightly
smaller radii than were obtained using the maximum of 35 m/s or the internally calculated value
(which can only be accessed by modifying ACE).
In the original validation of ACE, only expanded radii were compared with experiments. In the
current study, it was considered that this could be enhanced by comparison of expansion
velocities. This was because, based upon cloud radius, it initially appeared that most of the
cloud evolution could be explained by the explosion phase alone. However, comparison of
expansion velocities appeared to support the existence of the turbulent growth phase. The
experimental data were still limited somewhat by the timescales that could be achieved within
the bounds of the measuring scales. Predicted cloud expansion velocities showed different
behaviour to the general trend seen in the experiments but the overall timescales (and therefore
final radii) predicted by ACE were generally in line with those observed. This suggests that the
termination criteria used in ACE are reasonable.
Recommendations
The current version of ACE allows the user to change the turbulent velocity and the main effect
of doing this is to change the final cloud radius. There is no strong argument to change the
turbulent velocity from the default 10 m/s so it may be preferable to remove this option from
the user interface.
iv
1
INTRODUCTION
1. Until recently, HSE have employed the IRATE model to compute source terms arising from
instantaneous releases of superheated liquids. Following review of the IRATE model, (Tickle,
2012a) its use has been discontinued. In its place, HSE are considering the use of the ACE
(Airborne Concentration Estimate) model developed by WS Atkins (Gilham et al., 1997a,b).
Inputs to ACE are the material, its mass and storage conditions and ambient weather conditions.
The model then computes the formation of a vapour cloud assuming total loss of containment.
ACE outputs resultant cloud properties which can be used as the input to an atmospheric
dispersion model. The original ACE model assumed that all liquid droplets would be
completely evaporated due to entrainment of air into the cloud. A subsequent modification was
made (Shepherd and Deaves, 2000) so that a range of droplet sizes was considered, some of
which would evaporate and the remainder would form an aerosol or rain out into a pool. A
review of ACE has been carried out by Tickle (2012b) and the validation studies presented in
this report are based upon recommendations in Tickle (2012b).
2. ACE models the formation of a vapour cloud in two distinct stages:
1) An explosion phase, which models the initial flashing process.
2) A turbulent growth stage, describing further radial expansion and air entrainment. Cloud
properties can be obtained at the end of the turbulent growth stage.
3. The original ACE model incorporated a third stage to describe the slumping of the resulting
heavy gas cloud. This was subsequently changed so that output from the turbulent growth stage
could be input to an atmospheric dispersion model.
4. Partitioning the cloud development into separate phases is done as much for modelling purposes
as for physical reasons and the approach has been previously used, for example by Hardee and
Lee (1975). In their model, the initial explosion phase is complete after isentropic expansion to
atmospheric pressure and air entrainment does not begin until the second stage. In reality, the
process is continuous and the transition between stages not easily defined. However,
Makhviladze and Yakush (2004) fitted models to experimental data and suggested that some
distinction can be made between the two phases.
5. Shepherd and Deaves (2000) describe a limited validation of ACE but acknowledge that
validation is difficult due to the limited experimental data and the difficulty in its interpretation
in relation to the various stages of the release. Therefore, HSE wish to review the validation of
ACE in light of any new experimental data, or by more in-depth comparison with existing data
to determine whether they should use the ACE model as a source term model to replace the
IRATE model.
6. In this report, overall cloud radius predictions have been compared with those obtained from
experiment and these are reported in Section 2.4.1. In addition, an attempt has been made to
further the validation by comparing cloud expansion velocities. This was not previously carried
out for ACE and was considered an area where the validation could be enhanced. The
comparisons of expansion velocity are reported in Section 2.4.4.
1
2
2.1
VALIDATION
ORIGINAL VALIDATION
7. Gilham et al. (1997a,b) divided the validation into two parts: the explosion phase and the
turbulent growth stage. This was based upon the lack of experimental evidence relating to
turbulence within the cloud. Validation of the explosion sub-model was therefore carried out
against measurements of cloud size obtained from several sources as listed in Table 1.
Table 1 cloud size comparisons
Source
Maurer et al. (1977)
Pettitt (1990)
Hardee and Lee (1975)
Hess et al. (1974)
Schmidli et al. (1992)
Comparison
Cloud radius
Cloud radius
Cloud radius
Qualitative
Qualitative
8. Gilham et al. (1997b) state that direct validation of the turbulent growth stage would be
difficult, as the initial conditions are not fully characterised. Instead, overall predictions from
ACE were compared with CFD (Computational Fluid Dynamics) calculations which had been
themselves compared with the results from Pettitt (1990).
9. Much of the original validation was through comparisons of single values of cloud radius.
However, in the explosion phase, the equations used in ACE determine the evolution of the
cloud with time, until a characteristic radius has been reached. Additionally, many of the
experimental results are of cloud size as a function of time. Further value could therefore be
added to the existing validation by comparing the evolution of cloud size with time (Tickle,
2012b).
10. The turbulent growth sub model specifies the cloud radius in terms of a turbulent energy density
whose initial condition is a function of a “turbulent velocity,” uL. In the original model of
Gilham et al. (1997b), the turbulent velocity was calculated as a function of the flash fraction
and vapour/air densities:
(1)
where x is the flash fraction, ρa is the air density, ρva is the vapour density at ambient conditions
and ρsva is the saturation vapour density at the storage conditions. ACE was subsequently
changed so that the turbulent velocity had to be entered by a user with a default value of 10 m/s.
Determination of an appropriate value of the turbulent velocity is an aspect of the model that
would merit further investigation and is discussed further in Section 2.4.6.
2
2.2
EXPERIMENTAL DATA
11. In line with the widespread use and transport of pressurised liquefied gases, a relatively large
number of instantaneous release experiments have been performed. The energetic nature of
instantaneous flashing releases represents a considerable experimental challenge. Despite this,
there have been a number of examples ranging from small laboratory scale to full size tests
which have generated useful results. In some of these cases, the vapour clouds were ignited to
investigate the formation of fireballs and this does limit the use of those datasets. The smaller
laboratory scale tests tended to be extensively instrumented and therefore yielded more data,
particularly in relation to droplet size measurements.
2.2.1
Large scale tests
12. Maurer et al. (1977) carried out experiments involving instantaneous releases of propylene
(propene) from a range of tank sizes from 40 mm diameter × 180 mm length (0.124 kg) to 700
mm diameter × 2800 mm length (452 kg). The tanks were electrically heated before bursting
was initiated by either explosive or mechanical means. The actual pressures and temperatures
at which the tanks failed are not clear from the reference. Instead a temperature range is given
of 50 °C to 80 °C (corresponding to vapour pressures between 22 and 39 bar). However, the
text states that failure was then initiated at pressures of about 60 bar by a sharp edged lance or
small explosive charge. The resulting vapour clouds were subsequently ignited by electrical
means and some information on the formation of the vapour clouds prior to ignition is
presented. There is a large uncertainty over the pressure prior to failure as the reference is not
clear whether the 60 bar was achieved by further heating.
13. In the validation of the explosion sub-model, Gilham et al. (1997b) used the 40 mm (0.124 kg)
experiment for comparison, possibly as the final pressure is given in the figure. Useful
information is also available from the 452 kg experiment in the form of a graph of cloud radius
versus time and is reproduced in Figure 1. As shown in Figure 1, these releases were ignited at
approximately 0.5 s and therefore the progress of the vapour cloud is only available up until this
time.
3
Figure 1 452 kg propylene test results (from Maurer et al., 1977)
14. Hardee and Lee (1975) present results (Figure 2) from an experiment in which 421.8 kg of
propane was released and is therefore of very similar size to the experiments of Maurer et al.
(1977). The initial conditions were different from those in Maurer et al. (1977) in that the
propane was not preheated. The storage temperature, whilst not explicitly given in the text,
appears to be an ambient value of 294.4 K. A further difference with the Maurer et al. (1977)
experiments appears to be in the size of the aperture in the tank. Hardee and Lee (1975) do not
give a value other than stating that there was a “relatively small” tank opening and the results
show that the cloud radius develops over a period of 4 seconds. There is therefore some
uncertainty as to how this compares with the tank opening methods of Maurer et al. (1977).
4
Figure 2 421.8 kg propane test results (reproduced from Hardee and Lee, 1975)
15. Johnson and Pritchard (1991) carried out experiments in which vessels containing 1-2 tonnes of
butane or propane were heated then failed and ignited. These experiments were used by Shield
(1993) in deriving a model for predicting vapour cloud and fireball growth. The Shield model
for the initial explosion phase was adopted in principle in the development of ACE. The
drawback of the Johnson and Pritchard (1991) results is that they focus on the development of
the fireball and little information (aside from the original video footage) is available on the
initial stages of vapour cloud development.
16. More recently, Birk et al. (2007) exposed a series of 0.4 m3 and 2 m3 ASME (American Society
of Mechanical Engineers) code propane tanks to torch and pool fires. 60 tanks were tested and
20 of those resulted in catastrophic failure. The tests were videoed but the main interest of the
tests was the blast overpressure and no measurements were made of the cloud dimensions. The
authors do provide a description of the initial stages of the tank opening and this illustrates the
difficulty in predicting the shape of the vapour cloud given the uncertainty in the tank opening
mechanism.
2.2.2
Small scale tests
17. Small scale tests of flashing releases were undertaken by Bettis (1987), Pettitt (1990), Schmidli
et al. (1992) and Bettis and Jagger (1992). The aim of these small scale tests was to characterise
the source arising from flashing releases and therefore the tests were extensively instrumented.
The experiments of Bettis (1987) involved releases of up to 1 kg of Freon-11. These tests
differed somewhat from the other small scale tests in that the vessel was composed of two
halves clamped together by pneumatic rams. The vessel was then “failed” by pulling the two
halves apart, leading to a “ring-doughnut” rather than spherically shaped cloud. Pettitt (1990)
also studied Freon-11 releases but the vessels used were glass spheres. These were failed either
due to internal pressure by heating or by mechanical impact. Pettit (1990) experimented with
varying fill fractions and provides a series of plots of cloud front velocity versus time (Figure
3). Bettis and Jagger (1992) carried out similar experiments, but with quantities of Freon-11 up
5
to 10 litres contained in a 20 litre vessel. Further analysis of these tests was performed by
Webber et al. (1992) from which Figure 4 is taken.
Figure 3 Glass sphere Freon-11 test (from Pettitt, 1990)
Figure 4 Glass sphere Freon-11 test (from Webber et al., 1992)
18. Schmidli et al. (1992) employed shattering glass spheres as the release mechanism, for releases
of refrigerant 114, refrigerant 12 and propane. Figure 5, taken from Schmidli et al. (1992),
shows the increase of cloud radius with time for a 2 litre propane release. One of the main
advantages of the glass sphere experiments is that they closely represent the ideal instantaneous
release scenario.
6
19. Hasegawa and Sato (1977) also used glass spheres in their experiments using 0.31 to 6.2 kg of
n-pentane. The aim of the study was to investigate fireballs and therefore the vapour clouds
were ignited by a pilot flame. Little information is therefore available on the dimensions of the
vapour clouds prior to ignition.
Figure 5 Glass sphere propane test (from Schmidli et al., (1992)
2.3
MODIFICATION OF ACE OUTPUT
20. In its standard form, during the initial rapid expansion explosion phase, ACE does not compute
the evolution of the cloud with time. Instead, it arrives at an asymptotic radius, Rc, given as the
product of a constant, N and a length scale L:
(2)
However, the cloud radius as a function of time is given by Shield (1993) as:
(3)
Where t is the time and:
(4)
tP is a timescale given by:
(5)
21. During the turbulent growth phase, the cloud radius is computed iteratively as a function of time
even though only the final radius is reported by the code. For the validation exercise, the
source code of ACE was therefore modified to provide a radius as per Equation 3 and also for
each time step during the turbulent growth phase.
7
2.4
VALIDATION RESULTS
2.4.1
Comparison of absolute radii
2.4.1.1
Maurer et al. (1977)
22. Due to the uncertainty in the initial conditions in the experiments, the ACE runs were firstly
carried out at a preheat of 50 °C (22 bar vapour pressure). Three separate runs were undertaken
with different values of uL. The first two were user input as the default of 10 m/s and the
maximum of 35 m/s. The third was by modifying ACE to compute its own value via Equation
1. In all the runs, the fate of the liquid was set to “vaporisation” which would add the highest
quantity of material to the cloud. The ACE runs were all carried out using the default
omnidirectional release option and the results are shown in Figure 6. Results of the model
output in both the explosion and turbulent growth phases are plotted from zero initial time as
the model does not distinguish between the phases on a time basis. Also plotted in Figure 6 is
an estimate of the radius (of a sphere) obtained from an isentropic flashing to atmospheric
pressure.
23. In terms of the overall cloud radius, comparison with experiment is limited to the point that the
cloud was ignited at about 0.5 s (determined from Figure 1). In this region, the explosion sub
model gives relatively good agreement using uL = 10 m/s. Higher values of uL result in a greater
final cloud radius, with the value of uL computed using Equation 1 being similar to that using
35 m/s. Both the experiments and predictions show significant post-flashing growth. Increasing
the preheat to 80 °C results in a somewhat larger cloud radius (Figure 7) and changing the
liquid fate to rainout has almost no effect on the cloud radius (Figure 8). Inspection of the ACE
output file showed the run including vaporisation results in only a very slightly larger and
cooler cloud.
8
Figure 6 Comparison of ACE results with the data of Maurer et al. (1977) for different
values of turbulent velocity
Figure 7 Comparison of results for 50 °C and 80 °C preheat
9
Figure 8 Comparison of droplet fate options with the data of Maurer et al. (1977)
2.4.1.2
Hardee and Lee (1975)
24. The Hardee and Lee (1975) results are shown in Figure 9. As mentioned in Section 2.2.1, there
is some uncertainty over the timescale of the release. Nevertheless, a 10 m radius is reached
more quickly and a larger cloud radius is produced compared to the Maurer et al. (1977) tests,
for a similar liquid mass. In this case, ACE predicts a slightly smaller final radius, with the
higher value uL providing a better fit. Again, significant expansion appears to occur beyond the
simple isentropic flashing radius.
Figure 9 Comparison of ACE results with the data of Hardee and Lee (1975) for
different values of turbulent velocity
10
2.4.1.3
Pettitt (1990)
25. Pettitt (1990) performed a large number of tests with differing release masses and fracture
mechanisms. The test used for comparison with ACE is a 1 litre sphere with a 1 litre Freon-11
fill level. In this test, the sphere was fractured by external impact rather than internal pressure.
The results are shown in Figure 10. The predicted radius comparison in this case is again
limited by the fact that the measuring scale was limited to cloud radii less than 0.4 m (only
slightly larger than the isentropic expansion estimate). Again, almost no difference was
obtained using the droplet rainout versus vaporisation options (Figure 11).
Figure 10 Comparison of ACE results with the data of Pettitt (1990) for different values
of turbulent velocity
Figure 11 Comparison of droplet fate options with the data of Pettitt (1990)
11
2.4.1.4
Schmidli et al. (1992)
26. The only test for which radius data were available is the release of 2 litres of propane from a
glass sphere and the results are shown in Figure 12. Good agreement with this test was obtained
using ACE with uL = 10 m/s. However, the maximum radius obtained in the experiment may
again have been limited by the measuring scale. The experimental data in Figure 5 do show a
flattening around 10 ms suggesting that this is where post flashing begins. Little further
expansion is shown beyond that point.
Figure 12 Comparison of ACE results with the data of Schmidli et al. (1990)
2.4.1.5
Bettis and Jagger (1992)
27. Bettis and Jagger (1992) do not present any cloud radius results, but further analysis of the
series of tests is available in Webber et al. (1992), from which the variation of radius with time
was extracted. The measurements of radius were from video footage in the near field and from
temperature measurements beyond approximately 1 m. Good agreement with this test was
obtained with the default turbulent velocity setting, shown in Figure 13. The video footage for
this test shows a large quantity of liquid present in the cloud which did not fully vaporise. This
is not reflected in the ACE radius predictions which give similar results using either the rainout
or vaporisation options (Figure 14). The radial data collected from the video footage
(distinguished in Figure 4) appear to fall into the initial flashing phase with values less than
0.7 m, which is roughly that given by isentropic flashing.
12
Figure 13 Comparison of ACE results with the data of Bettis and Jagger (1992) for
different values of turbulent velocity
Figure 14 Comparison of droplet fate options with the data of Bettis and Jagger (1992)
2.4.2
Discussion of radius comparisons
28. A comparison of absolute radii shows that the ACE results are in reasonably good agreement
with the experiments. In the larger scale tests, predictions fall approximately within a factor of
two of the experiments. Results for the smaller scale tests indicate that the predicted rate of
cloud growth in the explosion phase compares well with the experiments. However, with the
exception of the Hardee and Lee (1975) data, comparison of the radii also appears to show that
most of the experimental data can be explained by the explosion phase alone. Care is needed in
the interpretation of these results as, in many cases, the measured maximum cloud size is
limited by the extent of the measurements. For example, the slope of cloud growth in the small
13
scale tests indicates that clouds were still growing when the limit of measurement had been
reached.
29. The differing results from the Hardee and Lee (1975) experiments may be due in part to the
non-catastrophic release mechanism. In a depressurisation through a relatively small hole the
expansion process is limited by the rate at which material can leave the vessel. In a rapid
depressurisation (such as a breaking glass sphere) the expansion process is limited by bubble
nucleation within the superheated liquid.
2.4.3
Non-dimensional comparisons
30. The substantially differing liquid masses and timescales between the large and small scale
releases means that model comparison on absolute terms may be of limited use. The value of
radius comparisons is also limited by the fact that the final radius could not be measured in
some of the experiments. Furthermore, the data do not reveal an obvious distinction between
the explosive expansion and turbulent growth stages. On this basis, it is difficult to justify the
existence of a turbulent growth model other than that this aspect has been observed in tests.
More insight may be gained using a non-dimensional approach and considering time and
expansion velocity. Makhviladze and Yakush (2004) present a number of experimental datasets
on a dimensionless time and cloud front velocity basis, as shown in Figure 15.
Figure 15 non dimensional expansion velocity with time (from Makhviladze and
Yakush, 2004)
31. The experimental data were collapsed on the basis of time and velocity quantities as follows. A
length scale, L*, was defined as:
(6)
Where M is the release mass, Rg is the gas constant and Ta and Pa are the ambient temperature
and pressure respectively. This length scale represents a cloud dimension based upon the
expansion of the mass of material to ambient conditions. A velocity scale, U* was defined as:
14
(7)
where hl is the liquid enthalpy and hv is the vapour enthalpy at initial (subscript 0) and ambient
(subscript a) conditions. xv is the mass fraction of vapour determined assuming isentropic
expansion:
(8)
where sl is the liquid entropy and sv is the vapour entropy again at initial (subscript 0) and
ambient (subscript a) conditions and can be obtained from tables (e.g. Braker and Mossman,
1980). The isentropic expansion assumes that there is no exchange of material with the
atmosphere so that during expansion, the initial liquid entropy is shared between the remaining
liquid and vapour. A time scale was then defined as:
(9)
32. Use of such a velocity scale had previously been visited by Webber et al. (1992) in comparing
the results of the Bettis and Jagger (1992) glass sphere tests. The authors noted that Equation 7
does not distinguish between the kinetic energy of the bulk expansion and that contained in
turbulence. Equation 7 will therefore tend to overestimate the expansion velocity.
33. Figure 16 shows the non-dimensional expansion velocity (U/U*) versus time (t/t*) for the
experimental data of Maurer et al. (1977), Hardee and Lee (1975) and Pettitt (1990). At early
times, the small scale results of Pettitt (1990) tend to give expansion velocities approaching that
predicted by Equation 7. Makhviladze and Yakush (2004) suggest that the transition from the
explosive expansion phase to a turbulent growth stage occurs at approximately t/t* = 1. With
this assumption, the data can be seen to span both regimes. However, the distinction is not
completely clear as the small scale results tend to sit mainly in the explosive expansion region
and the larger scale tests tend to sit in the turbulent growth region. As noted in Section 2.4.2, it
is possible that the limiting factor in the small scale tests is that the end of the measuring scale
was reached before the cloud had fully expanded. The length scale predicted by Equation 6
tends to be somewhat larger than the measured maximum cloud size for the Pettitt (1990) tests
and the Schmidli et al. (1992) tests.
34. The small scale test data are shown separately (for clarity) in Figure 17 as the points for the
Schmidli et al. (1992) and Bettis and Jagger (1992) tests do not collapse as neatly as for the
larger scale experiments. In these two tests, the initial stages of cloud growth are fairly linear
and it is possible that noise in the data becomes dominant.
15
Figure 16 experimental data from Maurer et al. (1977) and Hardee and Lee (1975)
presented in non-dimensional form. Results from small scale test by Pettit (1990) are
included for comparison
Figure 17 small scale test data in non-dimensional form
16
2.4.4
Results of expansion velocity comparison
2.4.4.1
Maurer et al. (1977)
35. Figure 18 shows the comparison of predicted expansion velocity with time against the
experimental values from Maurer et al. (1977). The ACE predictions for both the explosion and
turbulent growth phases have both been plotted from t = 0 to identify a transition from one
phase to the other. In terms of the overall cloud radius the time of this transition point is
meaningless as ACE computes the radius in the turbulent growth phase only as a function of the
cloud radius and uL at the end of the explosion phase. The experimental data give an initial
expansion velocity of roughly half that given by Equation 7 and the ACE predictions result in
roughly the same when using a calculated value of uL. Using the default value of uL of 10 m/s
results in a slower initial expansion. The ACE predictions, for both values of uL show a later
transition from the explosion to turbulent growth phases than that suggested by Makhviladze
and Yakush (2004). The extent of the measured cloud expansion is limited by ignition at 0.5
seconds, corresponding to t/t* of approximately 30.
Figure 18 comparison of ACE with data of Maurer et al. (1977) in non-dimensional
form
2.4.4.2
Hardee and Lee (1975)
36. The Hardee and Lee (1975) data shown in Figure 19 when normalised, show a slower initial
expansion than the Maurer et al. (1977) data and a longer timescale. This was due to the release
being measured over 4 seconds and the cloud not being ignited. For a relatively small opening,
as suggested in the text, one might expect a jetting effect to occur and give rise to different
cloud properties compared to a catastrophic loss of containment. This may explain, in part, why
the data appear to sit further into the turbulent growth region.
17
Figure 19 comparison of ACE with data of Hardee and Lee (1975) in non-dimensional
form
2.4.4.3
Pettitt (1990)
37. The measurements from Pettitt (1990) (Figure 20) are curtailed at a radius of 0.4 m. Up to this
point, the ACE predictions in the explosion region are in good agreement using a uL of 10 m/s.
Predictions using the calculated value of uL give a much more rapid expansion.
Figure 20 comparison of ACE with data of Pettitt (1990) in non-dimensional form
2.4.4.4
Bettis and Jagger (1992)
38. An interpretation of the Bettis and Jagger (1992) data is given in Webber et al. (1992) who
present both video (initial expansion) and temperature data (later stages) to capture the cloud
18
evolution. There is a large amount of scatter in the data but the extension to temperature array
data does provide evidence of the final cloud size. This is suggested in Figure 21 where a
longer time period was achieved. The ACE predictions are consistent in terms of the timescale,
but there is insufficient clarity to draw further conclusions.
Figure 21 comparison of ACE with data of Bettis and Jagger (1992) in non-dimensional
form
2.4.5
Termination criteria
39. The previous comparisons of expansion velocity are of single experiments and ACE runs. A
further useful comparison is to plot the ACE results in comparison to the whole experimental
dataset; this is shown in Figures 22-25. In these plots, the experimental data from Makhviladze
and Yakush (2004) have been plotted alongside Hardee and Lee (1975), Pettitt (1990), Bettis
and Jagger (1992) and Schmidli et al. (1992). Makhviladze and Yakush (2004) appear to
include the data of Maurer et al., (1977) in their plot so this has not been included separately.
Very approximately, the experimental data sit between the non-dimensional timescales of 0.1
and 100 (the final point in the Hardee and Lee (1975) data is 120). The ACE runs were all
carried out using the default turbulent velocity of 10 m/s and using the droplet vaporisation
option. Throughout the expansion, the ACE predictions do not exhibit the same behaviour as
seen in the experiments, with a much more abrupt reduction in velocity. However, in terms of
timescale, the termination criteria used by ACE appears to be not unreasonable. Other than the
prediction of the Maurer et al. (1977) test, which terminates at approximately t/t* = 200, the
ACE runs terminate close to t/t* = 100. The reason for the discrepancy is not clear.
19
Figure 22 ACE predictions of Maurer et al. (1977) plotted against all data
Figure 23 ACE predictions of Hardee and Lee (1975) plotted against all data
20
Figure 24 ACE predictions of Pettitt (1990) plotted against all data
Figure 25 ACE predictions of Bettis and Jagger (1992) plotted against all data
2.4.6
Discussion of expansion velocity comparisons
40. Comparison of expansion velocities and timescales shows that cloud growth cannot be
explained entirely by the initial rapid expansion to atmospheric pressure. On the basis of the
theoretical end of the explosion phase being at t/t*= 1, the experiments all appear to show that
significant expansion occurs after this point. The only exception to this is the data of Pettitt
(1990) in which the measurements were limited to a radius of 0.4 m. The large scale
21
experiments of Hardee and Lee (1975) and Bettis and Jagger (1992) are the only experiments
that appear to have captured the full extent of the cloud and this occurs at very approximately
t/t* = 100. It is worth noting that the ACE predictions terminate at a similar point in time.
41. Using a turbulent velocity computed by Equation 1 results in a higher initial expansion speed
and an earlier transition to the turbulent growth phase than would be obtained using the default
of 10 m/s. The higher initial expansion speed has no bearing on the cloud radius at the end of
the explosion phase. The main effect of altering the turbulent velocity is therefore to change the
final cloud radius. In terms of the expansion velocities and timescales, there is no conclusive
evidence which would support a move from the default value of 10 m/s, or the computed value
from Equation 1. However, using 10 m/s results in a slightly smaller cloud radius and therefore
more conservative predictions for flammable materials when subsequent dispersion is taken
into account.
22
3
CONCLUSIONS
42. The original validation of ACE was on the basis of cloud radius for the initial explosion phase
and against CFD predictions for the turbulent growth phase. Comparison of overall cloud radius
predictions against experiments at a range of scales showed that the ACE predictions were in
reasonable agreement. Particularly good agreement was obtained against the large scale tests of
Hardee and Lee (1975) and small scale tests of Bettis and Jagger (1992). For the other tests,
ACE produced much larger cloud radii. The main deficiency of absolute radius comparisons is
that it is not clear in many cases what the final cloud radius was; either the end of the measuring
scale was reached (e.g. Pettitt, 1990) or the cloud was ignited (Maurer et al., 1977). Using the
default turbulent velocity value of 10 m/s resulted in slightly smaller radii than was obtained
using the maximum of 35 m/s or the internally calculated value (which can only be accessed by
modifying ACE).
43. In the original validation of ACE, only expanded radii were compared with experiments. In the
current study, it was considered that this could be enhanced by comparison of expansion
velocities. This was because, based upon cloud radius, it initially appeared that most of the
cloud evolution could be explained by the explosion phase alone. However, comparison of
expansion velocities appeared to support the existence of the turbulent growth phase. The
experimental data were still limited somewhat by the timescales that could be achieved within
the bounds of the measuring scales. Predicted cloud expansion velocities showed different
behaviour to the general trend seen in the experiments but the overall timescales (and therefore
final radii) predicted by ACE were generally in line with those observed. This suggests that the
termination criteria used in ACE are reasonable.
23
4
REFERENCES
Bettis, R. J., (1987), “Two phase releases following rapid vessel failure,” PhD Thesis. South
Bank Polytechnic.
Bettis, R. J. and Jagger, S. F., (1992), “Some experimental aspects of transient releases of
pressurised liquefied gases,” IChemE symposium series No. 130, Major Hazards Onshore and
Offshore.
Birk, A. M., Davison, C. and Cunningham, M., (2007), “Blast overpressures from medium scale
BLEVE tests,” Journal of Loss Prevention in the Process Industries 20 pp 194–206.
Braker, W. and Mossman, A. L., (1980), Matheson Gas Data Book, 6th Edition.
Gilham, S., Moran, K.C. and Mitchell, B.H., (1997a), “Modelling of catastrophic flashing
releases of liquid: Phase I”, WS Atkins report number: WSA/RSU8000/64/1.
Gilham, S., Moran, K.C. and Mitchell, B.H., (1997b), “Modelling of catastrophic releases of
flashing liquids: Phase II”, WS Atkins report number: WSA/RSU8000/85/1.
Hardee, H.C. and Lee, D.O., (1975), “Expansion of clouds from pressurised liquids,” Accid.
Anal. and Prev., vol. 7, pp 91-102.
Hasegawa, K. and Sato, K., (1977), “Study on the fireballs following steam explosion of npentane,” proceedings of the second international symposium on loss prevention and safety
Promotion in the process industries, Heidelburg.
Hess, K., Hoffmann, W. and Stoeckel, A., (1974), “Propagation processes after the bursting of
tanks filled with liquid propane - Experiments and mathematical model,” Int. Loss Prevention
Symposium, The Hague, Netherlands, pp 227-234.
Johnson, D. M., and Pritchard, M. J., (1991), “Large scale experimental study of boiling liquid
expanding vapour explosions (BLEVEs),” MRS E 211, British Gas PLC, Session 3, paper 3.3.
Makhviladze, G. M. and Yakush, S. E., (2004), “Modelling the formation and combustion of
accidentally released fuel clouds,” IChemE symposium series No. 150.
Maurer, B., Hess, K., Giesbrecht H. and Leuckel, W., (1977), “Modelling of vapour cloud
dispersion and deflagration after bursting of tanks filled with liquefied Gas,” Int. Loss
Prevention Symposium, Heidelberg.
Pettitt, G.N., (1990), “Characterisation of two-phase releases,” PhD thesis, South Bank
Polytechnic.
Schmidli, Y., Yadigaroglu, G., and Banerjee, S, (1992), “Sudden releases of superheated
liquids,” HTD, vo1.l97, Two-Phase Flow and Heat Transfer, ASME.
Shepherd, A. H., Deaves, D. H., (2000), “Source Term Calculation for Flashing Releases Background to the Development of ACE,” WS Atkins report No. AM5233-R1.
Shield, S.R., (1993), “A model to predict the radiant transfer and blast hazards from LPG
BLEVES,” Vol. 89, AICHE Symposium Series.
24
Tickle, G. A., (2012a), “Review and Comments on the Models IRATE and CREATE”, GT
Science & Software Ltd report No. GTS/HSE/R001 draft A.
Tickle, G. A., (2012b), “Review and Comments on the Model ACE,” GT Science & Software
Ltd report No. GTS/HSE/R002 draft C.
Webber, D. M., Tickle, G. A., Wren, T and Kukkonen, J., (1992) “Mathematical modelling of
two-phase release phenomena in hazard analysis,” AEA Technology report SRD/HSE R584.
25
Published by the Health and Safety Executive
12/14
Health and Safety
Executive
Further validation of the ACE instantaneous
source model
ACE (Airborne Concentration Estimate) is a model originally
developed by WS Atkins on behalf of HSE for computing
source terms arising from instantaneous flashing releases.
The model has since been reviewed and enhanced through
a series of extensive tests that have been performed by
HSL. Inputs to ACE are the substance, its mass and storage
conditions and ambient weather conditions. The model then
computes the formation of a vapour cloud assuming total
loss of containment. This is done in two distinct stages; an
explosion phase, which models the initial flashing process and
a turbulent growth stage, describing further radial expansion
and air entrainment. The resulting cloud properties can be
used as the input to an atmospheric dispersion model. This
report describes an updated validation of ACE in which overall
cloud radius predictions have been compared with those
obtained from experiment. In addition, an attempt has been
made to further the validation by comparing cloud expansion
velocities. This was not previously carried out for ACE and was
considered an area where the validation could be enhanced.
This report and the work it describes were funded by the
Health and Safety Executive (HSE). Its contents, including
any opinions and/or conclusions expressed, are those of the
authors alone and do not necessarily reflect HSE policy.
RR1028
www.hse.gov.uk
Fly UP