SYMPOSIUM SERIES NO.

155 Hazards XXI # 2009 IChemE

CORRECTING THE PREDICTIONS BY BAKER-STREHLOW-TANG (BST) MODEL

FOR THE GROUND EFFECT

Yongfu Xu David Worthington and Adeyemi Oke

DNV Software, London

Simple models representing Vapour Cloud Explosions (VCEs) are widely used in risk assessment for
onshore and offshore process installations. Because of their simplicity, blast loads and the risks of a
large range of scenarios can be assessed quickly. The Baker-Strehlow-Tang (BST) model belongs to
a group of these models using blast curves to predict the consequences of VCEs. The blast curves are
usually developed based on modelling results of idealised cases and provide side-on overpressure and
impulse or positive duration as a function of distance from the explosion centre. The blast curves
developed for BST methodology were based on the predictions of spherical free air explosions
and its predictions need to be corrected when applied for explosions near to ground. This paper
presents a method to correct the predictions of the BST model and validation of this method.

KEYWORDS: Vapour cloud explosion (VCE), Baker-Strehlow-Tang model (BST), ground effect,
blast curves, overpressure and Phast Risk

INTRODUCTION the following for the blast effect of vessel ruptures near to
The Baker-Strehlow-Tang (BST) model belongs to a group the ground. The effective blast wave energy is given by:
of simple models using blast curves to predict the conse-
quences of VCEs. The other models in this group include Eex ¼ Asb Eav
TNO Wiekema model (Lees, 1996), TNO Multi-Energy
model (TNO, 1997) and the Congestion Area Assessment Where Eav is the available energy and Asb is the
model (CAM) (Puttock, 1995). The blast curves are ground reflection correction factor. In this approach, it is
usually developed based on modelling results of idealised assumed that all available energy will be converted into
cases and provide side-on overpressure and impulse or posi- blast wave energy and the presence of a reflecting ground
tive duration as a function of distance. The blast curves are surface is taken into account through the factor Asb. Asb
then used for assessing the consequences and risks of VCEs. equals one when the vessel is high in the air and two
Because of their simplicity, a large range of scenarios can be when the vessel is less than 15 degrees above the horizon
assessed quickly for QRA. as seen from the target.
The Baker-Strehlow model was first published at the Fitzgerald (2001) used a reflection factor of 2 for
28th Loss Prevention Symposium in 1994 (Baker et al, ground vapour cloud explosions. Correcting the explosion
1994) This methodology includes a flame speed table, energy has improved predictions of overpressure away
through which the flame speed is decided according to from the explosion source, however, the corrected BST
the confinement, congestion and cloud reactivity of the model still under-predicts overpressures of ground VCEs
obstructed regions, and a family of blast curves for overpres- significantly, particularly at the explosion source, as pre-
sure and impulse at flame Mach numbers covering deflagra- sented in the Fitzgerald paper.
tion to detonation. Since 1994, the flame speed table had Accurate predictions of peak overpressures of VCEs
been updated to include a confinement of 2.5D in 1997 are often sought after, especially for the design of process
(Baker et al., 1997), the blast curves were updated in 1999 installations (Huser, 2009). Simple models, such as BST
and the model was then renamed as Baker-Strehlow-Tang and Multi-energy models, are particularly useful at the
model (Tang et al., 1999), i.e. BST model. A new early stage of design because the details required by CFD
flame speed table was published in 2005 which gives models are often unavailable. A simple method is presented
flame speed relative to a fixed observer, instead of relative in this paper to correct the BST predictions for ground
to the moving gas as by the original flame speed table, effect. The results are validated against measurements
and has excluded 1D confinement (Pierorazio et al., 2005). and predictions using CFD, the TNO Multi-Energy model
The BST model used for this study employs the updated and the BST model using the ground reflection correction.
flame speed table and blast curves.
The blast curves developed for the BST methodology
were based on the predictions of spherical free air
GROUND CORRECTION METHOD
explosions. Corrections are required to count for the
Factors determining the consequence of a VCE include:
ground effect for explosions on or near to ground. The
current approach is to apply a ground reflection factor to . Fuel type and concentration
the explosion energy. TNO Yellow Book (1997) suggests . Ignition location and strength

318
SYMPOSIUM SERIES NO. 155 Hazards XXI # 2009 IChemE

. Confinement and venting given by:

. Congestion, i.e. blockage ratio and obstacle diameters
 0:33
The GAME project (Eggen, 1998) has derived corre- 3V
R1 ¼ ¼ 0:784V 0:33
lations relating these parameters to the peak overpressures 2p
of VCEs for two confinement geometries of low ignition
strength. The correlations are: Where V is the volume of flammable cloud trapped
For no confinement (open, 3D expansion); inside the obstructed regions.
Area of the venting surface (i.e. assuming the flamma-
ble cloud may be suitably represented by a hemisphere) is
 
VBR  LP 2:75 2:7 0:7
Pmax ¼ 0:84 SL D (1)
D S1 ¼ 3:86V 0:66

For the VCE in air, flame path length of the worst-

For confinement between parallel plates (2D expansion); case scenario would be normally radius of the sphere as:
 0:33
  3V
VBR  LP 2:25 2:7 0:7 R0 ¼ ¼ 0:623V 0:33
Pmax ¼ 3:38 SL D (2) 4p
D
Area of the venting surface is
With
S0 ¼ 4:88V 0:66
VBR volume blockage ratio of the obstructed regions
within an explosion source It is clear that the ground VCE has a longer flame path
D characteristic diameter of the obstructed regions length and a smaller venting surface, and this should lead to
Lp flame path length of the explosion a higher peak overpressure at the explosion source.
SL laminar burning velocity of the flammable cloud
Pmax peak overpressure at the explosion source and is
used to select blast curves for estimating blast CORRECTING THE PEAK OVERPRESSURE
effect. AT THE EXPLOSION SOURCE
It is proposed here to correct the peak overpressure for the
The locations of ignition and venting surfaces are not ground effect, combined with the correction of explosion
reflected in the correlations directly, but they should be con- energy as used by Fitzgerald (2001), and the ground correc-
sidered in the determination of flame path length for realistic tion factor is related to flame path length and area of the
predictions. venting surface as.
Figure 1 illustrates a simple ground VCE and a VCE
of the same volume in air. Assuming they have the same  a  0 b
R0 S
congestion, uniform obstacle configuration and an ignition Cf ¼ (3)
R0 S0
location of the worst-case scenario (i.e. central ignition),
the difference in peak overpressure between them is
With
mainly due to the ground effect. Flame path length of the
ground explosion would be the radius of the hemisphere
R0 radius of the truncated sphere representing an
explosion near to ground as shown in Figure 2
S0 venting area of the truncated sphere representing an
explosion
R0 radius of the sphere of equivalent volume to the VCE
(as shown in Figure 1)
S0 surface area of the sphere of equivalent volume to the
VCE
Cf ground correction factor

The effect of flame path length on peak overpressure is

assumed to be the same as in the GAME correlations.
Where a is given a value of 2.75 for 3D confinement and
2.25 for 2D confinement as shown in equations (1) & (2).
For 2.5D confinement of the BST methodology, a value of
Figure 1. Ground and air vapour cloud explosions (VCEs) 2.5 is interpolated for a.

319
SYMPOSIUM SERIES NO. 155 Hazards XXI # 2009 IChemE

. If the explosion source touches ground, the height of the

truncated sphere is adjusted so that the truncated sphere
has the same footprint as the explosion source until it is
zero, i.e. the explosion becomes a ground explosion, and
the radius is adjusted to enable its volume the same as
the explosion source.

CORRECTING THE EXPLOSION ENERGY

The explosion energy is corrected for the ground effect in
the same way as done by Fitzgerald (2001). A correction
factor of 2 is used for vapour cloud explosions near to or
on the ground. This may not be accurate in all cases, but
is considered to be conservative.
Figure 2. An elevated vapour cloud explosion and the
representation by a truncated sphere IMPLEMENTING THE GROUND CORRECTION
METHOD WITHIN THE BST METHODOLOGY
The steps to apply the ground correction method in the BST
Obstruction to venting by the ground should have also methodology are:
contributed to the increased peak overpressure of a ground
explosion, but it is considered to be less than that of flame 1. Determine the flame speed using the updated flame
path length for VCEs with a degree of confinement of 2D speed table (Pierorazio et al, 2005). No ground correc-
or higher. Also because no data was found to calibrate para- tion should be applied if the flame Mach number is
meter b confidently, the effect of venting surface is ignored already the highest Mach number in the table, which
here. The ground correction factor is simplified to: is 5.2.
2. Estimate the peak overpressure at the explosion source
 a using the correlation given by Tang and Baker (1999) as
R0
Cf ¼ (4)
R0
Pmax  P0 2:4Mf2
¼ (5)
The ground correction factor by equation (4) depends P0 1 þ Mf
on both volume and position of the VCE. For a VCE high
above the ground, the correction factor is: 3. Estimate a ground correction factor using equation (4)
and the correct the peak overpressure obtained from
Step 2.
R0 ¼ R0 (i.e. radius of the sphere as shown in Figure1) 4. Take the corrected peak overpressure into equation (5)
Cf ¼ 1 and determine the corrected flame speed. The corrected
flame Mach number is capped at 5.2.
This is the case of a free air explosion which had been 5. For consequence and risk calculations, the blast curves
simulated to derive the BST blast curves and has the are selected using the corrected flame speed and
minimum correction factor of 1. explosion energy is corrected using a correction factor
For a ground explosion, the correction factor is given in section 2.2.

R0 ¼ R1 (i.e. radius of the hemisphere as shown in Figure1)

VALIDATING THE METHOD
Cf ¼ 1:88 for 3D confinement Predictions using BST model with the proposed ground cor-
Cf ¼ 1:77 for 2.5D confinement rection method were validated against measurements and
Cf ¼ 1:68 for 2D confinement predictions of three models, i.e. BST model with a ground
reflection factor of 2, TNO multi-energy model and CFD
model AutoReaGas.
For an elevated vapour cloud explosion near to the
TNO Yellow Book (1997) has details of the Multi-
ground as shown in Figure 2, the centre and radius of the
energy model and the GAME and GAMES projects have
representing sphere (maybe truncated) are determined as
provided guidance on its applications. TNO Multi-energy
the follow:
model has been implemented in Phast Risk v6.60 (a soft-
. If the explosion source doesn’t touch the ground, the ware developed by DNV Software for risk assessment)
sphere locates at the centre of the explosion source and it was used to produce the results of Multi-Energy
and its radius is adjusted to ensure its volume the same model for this validation. BST model with given ground
as the explosion source. The sphere would be truncated reflection factors has also been implemented in Phast Risk
if its radius is larger than its height h. v6.60 and its results were included. AutoReaGas was used

320
 SYMPOSIUM SERIES NO. 155
Table 1. Input data for the multi-energy and BST models for the cases of BFETS3a, EMERGE and Shell Dear Park ethylene explosion (Fitzgerald, 2001)

Test setup BST input Multi-energy

input
Fuel
Flammable Degree of Congestion reactivity Explosion
Case Length (m) Width (m) Height (m) fuel VBR (%) Energy (J) confinement level level curve

BFETS3a_A 2 2 2 Methane 8.46 1.69E þ 10 2D High Low 6.908

(1 – 4)
BFETS3a_B 2 2 2 Methane 8.46 1.69E þ 10 2.5D High Low 6.908
(16 – 17,
19, 22)
BFETS3a_C 4 4 2 Methane 9.62 1.69E þ 10 2.5D High Low 7.452
(24 – 26,
29, 32)
BFETS3a_D 4 4 2 Methane 9.67 1.69E þ 10 2.5D High Low 7.557
(37 – 38)
BFETS3a_E 4 4 2 Methane 8.27 1.69E þ 10 2.5D High Low 6.976
(39 – 44)
EMERGE1 2 2 2 Methane 10 2.49E þ 07 3D High Low 7.051
(28 – 34)
321

EMERGE2 2 2 2 Propane 10 2.71E þ 07 3D High Medium 7.577

Hazards XXI
(40 – 42,
50 – )
EMERGE3 4 4 2 Methane 10 2.03E þ 08 3D High Low 7.378
(A1,4)
EMERGE4 4 4 2 Propane 10 2.15E þ 08 3D High Medium 7.904
(A2-3)
EMERGE5 4 4 2 Methane 4.8 2.03E þ 08 3D Medium Low 4.815
(F1, 3,
6 – 7)
EMERGE6 4 4 2 Propane 4.8 2.15E þ 08 3D Medium Medium 5.285
(F2,4 – 5)
EMERGE7 8 8 4 Methane 10 1.58E þ 09 3D High Low 8.139
(L1 – 2)
EMERGE8 8 8 4 Propane 10 1.69E þ 09 3D High Medium 8.586
(L3 – 4)
Shell Dear 50.6 25.3 15.2 Ethylene 10 1.36E þ 11 2.5D High High 10

# 2009 IChemE
Park with 19%
ethylene hydrogen
explosion
SYMPOSIUM SERIES NO. 155 Hazards XXI # 2009 IChemE

Table 2. Input data for the multi-energy and BST models for the GAMES test cases

The LNG terminal

The
The gas chemical OSR 1, 2 OSR 1, 3,
Parameter processing case plant OSR-1 OSR-2 &5 6&7
    
Length (m) 8.3
    
Width (m) 5.3
    
Height (m) 3.5
Degree of confinement 3D 3D 3D 3D 3D 3D
VBR 14.0% 13.0% 6.0% 4.0% 5.1% 4.8%
Congestion level High High Medium Medium Medium Medium
Flammable cloud Propane (4.2%) Methane LNG gas LNG gas LNG gas LNG gas
Cloud reactivity level Medium Low Low Low Low Low
Cloud laminar burning speed (m/s) 0.52 0.45 0.45 0.45 0.45 0.45
Typical diameter of all obstacles (m) 0.25 0.55 0.48 0.46 0.47 0.48
Flame path length (m) 4.2 15.9 12.1 8.9 14.6 15.9

The case consists of multiple obstructed regions and please refer to the GAMES report for details.

Radius of the equivalent hemisphere and central ignition is assumed.

in the GAMES study (Wercx et al, 1998) and its predictions . EMERGE. EMERGE tests were carried out by TNO,
for the two cases which have no measurements were also BG and CMR to investigate the effect of size, fuel reac-
used here for validation. tivity and congestion level on vapour cloud explosion
The four sets of test cases used for this validation are: (EMERGE, 1998).
. Shell Dear Park ethylene explosion (Fitzgerald, 2001).
. BFETS3a. BFETS2 was conducted by the Steel Con- . Test cases of the GAMES project. TNO Multi-energy
struction Institute for 1D vapour cloud explosions and model was evaluated against measurements of the case
BFETS3a was a follow-on project carried out by of Gas Processing and against AutoReaGas predictions
British Gas in 2000 to investigate gas explosions with for the other two cases, i.e. the Chemical Plant and the
the degree of confinement of 2D and 2.5D. Because LNG Terminal which have no measurements.
the new flame speed table is not recommended for 1D
explosion, the test results of BFETS2 were not used in Measurements of the first three sets were used in the
this study. Forty-five experiments were performed Fitzgerald study and the same data were used in this
in BFETS3a and 21 of them were included as in the study, Table 1 summarises the input data to the Multi-
Fitzgerald study (Fitzgerald, 2001). energy and BST models. Table 2 gives the input data for

Figure 3. Comparing the overpressure predictions of BST and multi-energy models against measurements and CFD predictions
(all cases)

322
SYMPOSIUM SERIES NO. 155 Hazards XXI # 2009 IChemE

Figure 4. Comparing the overpressure predictions of BST and multi-energy models against measurements (selected cases)

the GAMES test cases. The Chemical Plant and the LNG measurements and CFD predictions of all the test cases
Terminal consist of multiple obstructed regions and their listed above. The average errors are 278% and 230% for
details are not given in this paper and can be found in the BST model with a ground reflection factor of 2 and Multi-
GAMES report. energy model, and are consistent with the findings by
Fitzgerald (2001), which reported an average error of
RESULTS AND DISCUSSIONS 275% and 223% for BST and Multi-energy model respect-
BST model is a simple vapor cloud explosion model using ively. The small differences between the two studies are due
blast curves derived from modeling data of idealized cases to two reasons: (1) Figure 3 has included the GAMES test
and it best suits for VCEs in obstructed regions with cases and has excluded the BFETS2 cases, which are 1D
obstacles distributed uniformly in all directions. In reality, gas explosions and the updated flame speed table is not
great variations always exist in obstacle size and distri- recommended for them, but were included in the Fitzgerald
bution. The effectiveness of the model for real scenarios is study; (2) The Fitzgerald study applied an energy efficiency
better evaluated in a wide range of cases. Figure 3 compares of 20% and 50% for explosions with peak overpress-
the predictions of BST and Multi-energy models against the ure less than 0.5 bar and 1 bar respectively for the

Figure 5. Comparing the overpressures along a transect crossing the explosion source (BFETS3a_B)

323
SYMPOSIUM SERIES NO. 155 Hazards XXI # 2009 IChemE

Figure 6. Comparing the overpressures along a transect crossing the explosion source (The gas processing case of the
GAMES study)

Multi-energy model, and no energy efficiency was applied (i.e. EMERGE cases 1–6) and explosions with overpres-
in this study. sures higher than 10psi were excluded and only test cases
BST model with the ground correction method having measurements both in and outside the explosion
described above has improved the overpressure predictions. sources were included. Applying similar selection criteria,
The error of all cases has improved to 263%, as shown in Figure 4 shows comparison of the results by BST and
Figure 3, from 278% for the BST model with a ground Multi-energy models against measurements. Improved
reflection factor of 2. However, the under-predictions by predictions are achieved by all models. The error of
BST model are still quite significant. BST model with the ground correction method has
In the Fitzgerald study, the BST model produced reduced to below 249%, and this is significant better than
better predictions for more realistic scenarios. To reach 269% achieved by BST model with a ground reflection
the conclusion, test cases with small obstructed regions factor of 2.

Figure 7. Comparing the predicted positive duration along a transect crossing the explosion source (EMERGE6)

324
SYMPOSIUM SERIES NO. 155 Hazards XXI # 2009 IChemE

Figures 5–6 compares the predicted overpressures Baker, Q.A., Doolittle, C.M, Fitzgerald, G.A. and Tang, M.J.,
against measurements of two selected cases. Figures 5 1997, Recent development in the Baker-Strehlow VCE
shows improved predictions of overpressure by the BST analysis methodology, 31st Annual Loss Prevention
model with equation (4) for the ground effect. However Symposium.
Figure 6 shows the results of BST method with a ground Eggen, J.B.M.M., 1998, GAME: Development of guidance for
reflection factor of 2 in good agreement with the measure- the application of the multi-energy method, TNO Prins
ments for the Gas Processing case of the GAMES study, Maurits Laboratory, ISBN 0717616517.
the ground correction method has produced conservative EMERGE, 1998, Extended Modelling and Experimental
predictions at all measurement locations, but less conserva- Research into Gas Explosions, Final Summary Report for
tive than the Multi-energy model. No measurements were the project EMGERGE, CEC Contract EV5V-CT93-0274.
available to validate the predicted positive durations or 1998.
impulses. Figure 7 compares the predicted positive dur- Fitzgerald, G., 2001, A comparison of simple vapour cloud
ations for the EMERGE6 case by BST and Multi-energy explosion prediction methodologies, Second Annual
models. Symposium on “Beyond Regulatory Compliance: Making
Safety Second Nature” Mary Kay O’Conner Process
Safety Centre, Texas T&M University.
CONCLUSIONS Huser, A., Foyn, T. and Skottene, M., 2009, A CFD based
A correction method has been developed for the Baker- approach to the correlation of maximum explosion overpres-
Strehlow-Tang (BST) methodology for the ground effect sure to process plant parameters, Journal of Loss Prevention
of vapour cloud explosions. This method is based on the in the Process Industrials, 22, 324– 33.
GAME correlations for the Multi-energy model and takes Lees, F.P., 1996, Loss Prevention in the Process Industries,
into consideration the change in flame path length for Second edition, Butterworth Heinemann, ISBN
explosions on or near to the ground. Predictions using 0750615478.
BST model with this ground correction method have been Pierorazio, A.J., Thomas, J.K, Baker, Q.A. and Ketchum, D.E,
validated against measurements and predictions of other 2005, An update to the Baker-Strehlow-Tang vapour cloud
models for a wide range of test cases. explosion prediction methodology flame speed table,
Improved predictions were achieved using this Process Safety Progress, 24, 59 – 65.
ground correction method. The error of BST model with Puttock, J.S., 1995, Fuel Gas Explosion Guidelines - the Con-
the ground correction method have reduced to below gestion Assessment Method, Presented at the 2nd European
249%, and this is significant better than 269% achieved Conference on Major Hazards On- and Off-shore, Manche-
by BST model with only a ground reflection factor of 2. ster, England, October.
In spite of the improved predictions using the Tang, M.J. and Baker, Q.A., 1999, A new set of blast curves
presented correction method the predicted results are still from vapour cloud explosions, Process Safety Progress, 18,
significantly below the experimental results. 235–240.
Wercx, W.P.M., van den Berg, A.C. and van Leeuven, D. 1998,
Application of correlations to quantify the source strength of
REFERENCES vapour cloud explosions in realistic situations, Final report
Baker, Q.A. Tang, M.J., Scheier, E. and Silva, G.J., 1994, for the project: ‘GAMES’. TNO report PML 1998-C53
Vapour cloud explosion analysis, 28th Loss Prevention TNO Yellow Book, 1997. Methods for the calculation of
Symposium, AIChE. physical effect, CPR 14E, Part 2.

325