Journal of Loss Prevention in the Process Industries xxx (2015) 1e17

Contents lists available at ScienceDirect

Journal of Loss Prevention in the Process Industries

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

Estimation of explosion loading on small and medium sized

equipment from CFD simulations
Olav R. Hansen a, *, Malte T. Kjellander a, Remi Martini a, Jan A. Pappas b
a
Lloyd's Register Consulting, Bergen, Norway
b
Lloyd's Register Consulting, Sandvika, Norway

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

Article history: Explosion studies for design purposes are performed on daily basis among safety consultants all over the
Received 16 May 2015 world. For oil and gas facilities offshore, and often onshore, the computational fluid dynamics (CFD) tool
Received in revised form FLACS is usually applied, while others use simple blast curve formulations, like the TNO-Multi Energy
5 December 2015
Method. The purpose of the explosion studies is usually to give guidance on required design strength of
Accepted 6 December 2015
Available online xxx
equipment, piping, blast walls or buildings during design, or to verify a chosen design. One key element
is to translate the results from an explosion simulation into actual forces on equipment. For CFD studies
loads on large objects can usually be well estimated by reporting differential pressures across the objects.
Keywords:
CFD explosion modeling
For objects with key dimensions less than 2e3 grid cells (typically ~1m-2m), and in particular less than
FLACS one grid cell, this approach is not feasible. Industry guidance exists on how to estimate explosion loads
Explosion loading on piping and smaller equipment using a drag force formulation. This study demonstrates how the
Drag current guidance may lead to too low predicted explosion loads onto equipment. More precise methods
Pressure for load prediction onto piping, small and medium sized equipment are thereafter proposed and
evaluated.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction
The purpose of an explosion study is usually to give guidance on
required design strength or to confirm actual design for a facility or
nearby structures. Explosion studies are performed in a number of
ways. For offshore oil and gas installations a common approach
(NORSOK Z-013, 2010, and ISO, 19901-3, 2010) will be to demon-
strate that the installation can survive all explosion accidents with a
return frequency higher than 10
4/year, and to demonstrate this
several hundred CFD dispersion and explosion calculations are
usually performed (Hansen et al., 1999 and Hansen et al., 2013).
Some safety standards (e.g. API RP-752, 2009) require that build-
ings shall be designed to withstand a credible worst-case explosion,
and simple blast curves estimating free-field blast strength (pres-
sure, duration and impulse) as function of distance are expected
used to estimate the loads.
Regardless of the method and approach used, there is a need to
estimate the explosion loads onto piping, equipment, walls and
structures, and the conclusions from an explosion study will be
influenced by the way this is done. In a transient flow field during
an explosion pressure differences will build up around objects, and
by integrating the pressure over the surface of an object, a good
estimate of the explosion loading (forces onto the object) can be
obtained. To estimate the explosion load based on differential
pressures is thus the approach recommended for objects which are
properly resolved on the simulation grid in a CFD-simulation, like
blast walls, large objects and decks. Good explosion load estimates
from differential pressures can be expected for objects at least 2e3
grid cells across, while typical grid cell sizes for explosion studies
may be 0.5 me1.0 m. For smaller objects estimates of differential
pressures will be less accurate, and for objects with diameter less
than 1.0e1.5 grid cells, it may not even be possible to extract a
differential pressure from the simulation. For such objects the load
may be estimated from a general drag formula (Sand, 1999):

* Corresponding author.
E-mail address: olav.hansen@lr.org (O.R. Hansen).

http://dx.doi.org/10.1016/j.jlp.2015.12.004
0950-4230/© 2015 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Hansen, O.R., et al., Estimation of explosion loading on small and medium sized equipment from CFD
simulations, Journal of Loss Prevention in the Process Industries (2015), http://dx.doi.org/10.1016/j.jlp.2015.12.004
2 O.R. Hansen et al. / Journal of Loss Prevention in the Process Industries xxx (2015) 1e17

prioritization.
Nomenclature Load estimates for pipes and smaller equipment are usually only
considering the form drag, which is the first term of (1). This is the
API American Petroleum Institute (www.api.com) only component mentioned in (BFETS Interim Guidance Notes,
BFETS Blast and Fire Engineering for Topside Structures 1992), while in (FABIG TN-08) all terms of (1) are mentioned, but
(Joint Industry Project, 1990e1998) it is suggested that form drag using conservative drag coefficients
CFD Computational Fluid Dynamics from (Baker et al., 1983) may be the best way to describe loading for
DDT Deflagration to detonation transition objects with a diameter up to 1.0 m. In the Gas Explosion Engi-
DLM Direction Load Measurement neering Handbook (J.Czujko, 2001) a simulation shows a good
FABIG Fire and Blast Information Group (www.fabig.com) correlation between form drag and loading for a simple example
FLACS CFD-software for explosion modeling (www. case.
gexcon.com) With increasing size of object term two (inertia) and term four
FLNG Floating Liquified Natural Gas vessel (static pressure difference in the flow field) of Equation (1) will
FPSO Floating Production, Storage and Offloading vessel gradually become more important, and it is a question to what
HSE UK Health and Safety Executive (www.hse.gov.uk) degree the FABIG TN-08 guidance is valid with increasing object
ISO International Organiazation for Standardization sizes.
(www.iso.org) The guidance from FABIG TN-08 on how to estimate loads on
NORSOK Standards to ensure competiveness on the objects based on CFD-explosion calculations is as follows:
Norwegian Continental Shelf (www.standard.no)
PDF Pressure distribution function

(1)

Object size: Proposed method to estimate loads:

Here the first term is the form drag, the second term is an inertia
term (combined object buoyancy term in an accelerated flow-field/
added mass), third term has to do with density changes due to
combustion, fourth term is differential pressure in the flow field
and the last term is a hydro-elastic term giving potentially
increased drag force due to oscillations.
An accurate estimation of explosion loads is important. One
should be sure that the load estimates are representative or slightly
conservative, but on the other hand excessive conservatism is not
optimal as this gives a wrong, non-optimal resource allocation and
For multiple object situations, it is proposed to evaluate pressure
loss across the group of objects to estimate the loading.
FABIG TN-08 also proposes a relation between explosion pres-
sure and drag loads. This relation was taken from (Yasseri, 2002), but
is later also included in tabulated form in DNV recommended
practice (DNV-RP-D101, 2008 and DNV OS-A101, 2014). This way it is
possible to estimate drag loads on piping and other small equipment
from explosion studies where only overpressure is known.
The purpose of the study presented is to evaluate the accuracy of
the currently used methods for explosion load predictions relevant

Please cite this article in press as: Hansen, O.R., et al., Estimation of explosion loading on small and medium sized equipment from CFD
simulations, Journal of Loss Prevention in the Process Industries (2015), http://dx.doi.org/10.1016/j.jlp.2015.12.004
 O.R. Hansen et al. / Journal of Loss Prevention in the Process Industries xxx (2015) 1e17
to small and medium sized objects, and to propose improved
methods. To achieve this a number of FLACS explosion simulations
have been performed, in which explosion loads extracted using
differential pressure are compared to explosion loads estimated
using drag formulations.
2. Description of FLACS simulations
In the evaluations a known experiment is simulated, this is Test
4 from the HSE supported Phase 3A test series (see e.g. Hansen and
Johnson, 2015). This is an end ignited, low confined explosion in the
DNV GL Spadeadam 2600 m3 rig (28 m  12 m  8 m) with a near
stoichiometric mixture of air and natural gas (91.7% methane, 7%
ethane and 1.3% propane). In order to evaluate explosion loading
inside and outside the flame zone for a range of different condi-
tions, the following scenario variations have been modeled (the
maximum pressure listed is the highest reported at 35 monitor
points inside the explosion simulation).
Inside and outside the flame zone a number of rectangular ob-
jects (targets) were defined with characteristic dimensions from
0.5 m to 2.0 m, see Fig. 1 and Table 1. The numerical grid resolution
is 0.5 m in the entire domain, this to ensure that all targets are fully
on-grid and also that simulation results are accurate. About 1e1.5 m
to the side or above each target a pressure/drag/density sensor is
located from which free-field pressures and flow parameters are
reported. Average wall pressures from FLACS pressure panels are
also reported on each of the six object faces. The following targets
are defined:
In Fig. 2 maximum pressure field (left) and blast pressures
moments after flame exit (right) can be seen for the strong explo-
sion, showing pressure patterns around objects.
For all targets the following variable values are reported:
1 Transient average pressure (panel) on all 6 faces of the targets,
so that differential pressures can be extracted properly.
2 At free-field monitor points near target (~2e3 grid cells to the
side) the “side-on” pressure and drag are reported, as well as
directional drag and density.
The selected simulation cases and the various objects evaluated
for load monitoring include a wide range of loading cases of very
different character. This includes explosion scenarios of very
different strength, objects inside the explosion at various locations,
Please cite this article in press as: Hansen, O.R., et al., Estimation of explosion loading on small and medium sized equipment from CFD
simulations, Journal of Loss Prevention in the Process Industries (2015), http://dx.doi.org/10.1016/j.jlp.2015.12.004
3
and outside the explosion in different directions. Loads will include
sharp shock-waves, weaker blast waves, explosion wind or any
combination of these. If a load prediction method can handle all the
cases evaluated in this study with an acceptable accuracy, the
method should have a good potential to handle any loading sce-
nario onto single objects seen in gas explosion studies.
3. Evaluation of current industry guidance (FABIG TN-08/
DNV-RP-D101)
From the simulations performed and loading objects monitored,
current methods to estimate explosion loading have been evalu-
ated. The reported pressure and drag values at various free-field
monitor points have been used to estimate the explosion loading
on a selection of box shaped objects (1 m)3 from the simulations.
This object size is within the range of object sizes (up to 1 m
diameter) for which FABIG TN-08 considers form drag with a
conservative Baker drag coefficient to be the likely best way to
estimate loads. The following often used approaches will be
compared.
Method 1: Drag from Pressure (DfP)
In studies where only pressure predictions exist, either from
CFD, or from simpler phenomenological tools or blast curves, a
typical approach to estimate loading on piping and smaller objects
is to convert the pressure into a drag load. For this conversion the
recommended drag/pressure relation in FABIG TN-08 and DNV RP-
D101 is often applied, it should be mentioned that the latter
guidance recommends as a first approximation to conservatively
use Drag ¼ Pressure/3 for early phase studies. Conservative Baker-
drag coefficients are recommended in FABIG TN-08, and for box
shape we interpret that Cd ¼ 2.0 can be used, even if some studies
may conclude Cd ¼ 1.05 will be the best choice based on the same
guidance. For pipes Cd ¼ 1.2 is recommended by FABIG, while DNV-
RP-D101 suggests Cd ¼ 1.0 with reference to API-RP-2FB (API,
2006).
Method 2: Drag from CFD-simulations (Drag)
For this approach the same load calculation will be performed,
but instead of estimating the drag from pressures, it is reported
from the CFD-simulation. For the load estimate the same drag
4 O.R. Hansen et al. / Journal of Loss Prevention in the Process Industries xxx (2015) 1e17

Fig. 1. Array of targets (North is up).

coefficient (Cd ¼ 2.0) is used as for Method 1. studies using CFD as well as for simpler methods.

Method 3: Predicted static pressure (P) Method 4: Estimated reflected pressure (2P)

Often the side-on explosion pressure predicted nearby is used as In the far field of the explosion shock waves hitting an object
a first estimate of differential pressure load on an object, as the will be reflected leading to a higher pressure at the front of the
pressure distribution is the most common output from explosion object. As a first estimate a reflection coefficient of two is used for
shock waves.

Method 5: Direct load measurement (DLM)

Table 1
List of targets with positions, shape and dimensions.
As the last method considered as the correct reference for the
List of targets
other methods direct load measurements are reported. Transient
Name Only informative Used for load estimate Shape averaged panel overpressures for all 6 faces of the object are re-
Object position (lower Dimensions (m) ported and a transient resultant force is calculated.
corner) All forces in this section will be reported as pressures (force per
X Y Z XDIM YDIM ZDIM area), and since the projected front area of the targets is 1 m2, the
pressure unit kPa could be replaced by kN to extract the force.
BP1 33.5 5.5 0 1 1 1 B
BP2 39.5 5.5 0 1 1 1 B
As a first evaluation of Method 1 (DfP-i.e. drag estimated from
BP3 51.5 5.5 0 1 1 1 B pressures) maximum drag and maximum static pressure pairs re-
BP4 75.5 5.5 0 1 1 1 B ported from the weak explosion (simulation 400010), the strong
BP5 47 25 0 1 1 1 B explosion (simulation 400101) and the simulated deflagration-to-
BP6 61 39 0 1 1 1 B
detonation transition (simulation 400200) are plotted and
BP7 13.5 17.5 0 1 1 1 B
BP8 13.5 23.5 0 1 1 1 B compared to the drag/pressure relation of FABIG TN-08/DNV RP-
BP9 13.5 35.5 0 1 1 1 B D101, see Fig. 3. The first plot shows monitors located inside the
BP10 13.5 59.5 0 1 1 1 B flame zone, i.e. the 35 pressure monitors originally used to report
BP11
22 25 0 1 1 1 B
pressures in the experiment (located near the floor, decks, walls,
T1 39.5 10.5 3 3 0.5 0.5 BX
T2 59 3 4 2 0.5 2 B
etc.) as well as 12 monitors placed in the vicinity of the six targets
T3 41 2.5 3 0.5 3 0.5 BY placed inside the test rig in this simulation study. The second plot
T4 60 9 4 0.5 2 2 B shows all external monitors placed in the free-field in the vicinity of
T5 45 14 3 2 2 2 B the external targets outside the flame zone.
I1 1.5 9.5 1.5 1 1 1 B
Looking at the plots in Fig. 3 it should be evident that the FABIG/
I2 2 9 5.5 1 1 1 B
I3 21 10 5 1 1 1 B DNV guidance to estimate drag as function of explosion pressure
I4 26 2 5 1 1 1 B can be used in a non-conservative way. For the sensor locations
I5 26 1.5 1.5 1 1 1 B inside the explosion the relation can, with optimistic eyes, be said
I6 11 2 5 1 1 1 B to give a reasonable curve fit, but the majority of drag values re-
W1 27
10 0 1 1 1 B
W2 39
10 0 1 1 1 B
ported from the simulation are much higher than found from the
R1
6.5 5.5 0 1 1 1 B proposed relation. For the purpose it is intended used, and is being
R2
12.5 5.5 0 1 1 1 B used in the industry, the relation is thus questionable. Even the

Please cite this article in press as: Hansen, O.R., et al., Estimation of explosion loading on small and medium sized equipment from CFD
simulations, Journal of Loss Prevention in the Process Industries (2015), http://dx.doi.org/10.1016/j.jlp.2015.12.004
 O.R. Hansen et al. / Journal of Loss Prevention in the Process Industries xxx (2015) 1e17 5

Fig. 2. Example of maximum pressure pattern (left), and instantaneous pressure pattern just after flame exit (right) for the strong explosion simulation using guidelines for
improved blast predictions. Each target is monitored with 6 panels for direct load measurement (DLM) and a free-field monitor point 2e3 grid cells away.

somewhat more conservative approach using Drag ¼ Pressure/3 up change this conclusion much. For the external free-field monitors,
to a pressure of 200 kPa suggested by DNV RP-D101 does not distributed the way it is typically done in explosion studies, it is

Fig. 3. Relation between maximum overpressure and maximum dynamic pressure (drag) for monitors located inside (upper plot) and outside the explosion (lower plot).

Please cite this article in press as: Hansen, O.R., et al., Estimation of explosion loading on small and medium sized equipment from CFD
simulations, Journal of Loss Prevention in the Process Industries (2015), http://dx.doi.org/10.1016/j.jlp.2015.12.004
6 O.R. Hansen et al. / Journal of Loss Prevention in the Process Industries xxx (2015) 1e17

clear that the proposed relation gives much lower drag values
(typically a factor of two) than observed through the entire pres-
sure range for all three explosion strengths.
While overpressures tend to be higher in corners and on solid
objects, and drag will be higher for locations in the free flow (where
static pressures may be lower), a potentially conservative approach
would be to pick the highest pressure from any monitor (located at
walls and decks) to estimate maximum free-flow drag loads. For
the strong explosion the maximum reported internal pressure is
250 kPa, which corresponds to a drag load of 95 kPa, which is
representative for the highest drag reported for this scenario
(89 kPa near BP1 6 m East of rig). For the weak explosion with
maximum pressure 52 kPa, corresponding to 8 kPa drag according
to FABIG/DNV relation, a maximum drag of 11.5 kPa was found in
the simulation. To remain conservative for this scenario the sug-
gested simplification of using drag as 1/3 of pressure is required. For
the DDT scenario one monitor reported a pressure as high as 3 MPa
(1.7 MPa was reported at the same monitor in the actual experi-
ment). The maximum drag reported was 722 kPa 6 m East of rig exit
near BP1, just outside the detonating gas pocket. This drag load was
as high as the static pressure reported at the same location
(721 kPa), but of very short duration (3 ms). The estimated direct
load measurement prediction for nearby BP1 was 3104 kPa (31 bar).
For the comparison of load estimate methods explosions of
three strengths will be compared for 13 external targets on the
ground BP1-BP11 (6 m, 12 m, 24 m and 48 m East, 24 m and 43 m
NE, 6 m, 12 m, 24 m and 48 m North and 25 m NW), as well as R1
and R2 6 m and 12 m to the West, and the six internal targets I1eI6.
In the upper table of Table 2 shows the load estimates for the strong
explosion, both simulated using standard FLACS guidelines
(400100) and using guidelines for improved far-field blast pre-
dictions (400101) proposed by (Hansen and Johnson, 2015). Below
a similar table is shown for the weak explosion (400010) and the
strong explosion scenario assuming DDT (400200).

Table 2
Evaluation of various methods for explosion load prediction for strong explosion simulated using standard FLACS guidelines and improved guidelines for blast predictions
(upper table), and weak explosion and strong explosion with DDT (lower table).

Estimated load kPa (¼kN) Strong explosion (400100) Strong explosion (400101)

Standard FLACS guidelines Guidelines for improved blast predictions

Object Position 1-DfP 2-Drag 3-P 4-2P 5-DLM 1-DfP 2-Drag 3-P 4-2P 5-DLM

BP1 6mE 64 177 131 262 206 64 177 131 262 206
BP2 12 m E 42 85 100 201 149 42 87 102 205 152
BP3 24 m E 19 27 58 116 84 25 39 71 143 127
BP4 48 m E 5 5 24 49 30 8 8 33 66 54
BP5 24 m NE 10 13 39 78 46 12 16 45 89 62
BP6 43 m NE 4 3 20 40 22 6 5 27 54 39
BP7 6mN 6 22 25 51 28 6 22 25 51 28
BP8 12 m N 3 8 17 34 21 3 8 17 34 23
BP9 24 m N 2 2 11 23 16 2 2 12 23 18
BP10 48 m N 2 1 11 23 12 2 2 10 20 10
BP11 25 m NW 1 1 6 13 11 1 1 6 13 12
R1 6mW 3 5 13 27 17 3 5 13 27 17
R2 12 m W 2 6 12 23 16 2 6 12 23 16
External Geometric mean 16% 24% 66% 132% 86% 17% 28% 70% 140% 100%
I1 LD NW 4 2 18 35 5 4 2 18 35 5
I2 UD NW 4 15 18 35 10 4 15 18 35 10
I3 UD NE 36 61 90 180 68 36 61 90 180 68
I4 UD SE 57 63 122 243 72 57 63 122 243 72
I5 LD SE 62 70 131 262 83 62 70 131 262 83
I6 UD S 12 17 45 90 27 12 17 45 90 27
Internal Geometric mean 58% 79% 184% 368% 100% 58% 79% 184% 368% 100%

Estimated load kPa (¼kN) Weak explosion (400010) Strong explosion DDT at flame exit Experiment Experiment
(400200)

Object Position 1-DfP 2-Drag 3-P 4-2P 5-DLM 1-DfP 2-Drag 3-P 4-2P 5-DLM 3-P (Max) 3P (0.1ms time averaged)

BP1 6mE 8 23 32 64 35 1100 1443 721 1442 3104 Sensor failed Sensor failed
BP2 12 m E 5 9 23 47 24 220 429 275 550 610 389 294
BP3 24 m E 3 3 14 28 15 58 108 123 247 276 192 184
BP4 48 m E 3 1 13 26 13 10 12 40 80 70 45 42
BP5 24 m NE 2 1 10 20 11 46 71 106 211 221 Sensor failed Sensor failed
BP6 43 m NE 2 0 10 20 12 12 12 43 86 83 44 40
BP7 6mN 2 9 12 24 12 17 92 54 109 103 Flame influenced Flame influenced
BP8 12 m N 2 2 8 16 11 18 60 57 113 223 Sensor failed Sensor failed
BP9 24 m N 1 0 6 13 8 10 20 41 82 60 40 39
BP10 48 m N 1 0 6 12 6 4 6 21 42 33 27 26
BP11 25 m NW 1 0 5 11 5 5 9 22 44 37 19 18
R1 6mW 2 2 8 15 9 5 17 23 46 25
R2 12 m W 1 2 6 12 9 6 10 25 50 36 sim/exp
External Geometric mean 18% 13% 88% 175% 100% 17% 34% 51% 101% 100% 87% 94%
I1 LD NW 2 1 10 20 2 10 6 38 76 23
I2 UD NW 2 7 10 20 5 6 22 28 56 25
I3 UD NE 7 18 31 62 20 56 161 120 239 209
I4 UD SE 9 18 39 78 20 510 159 452 904 1086
I5 LD SE 9 17 38 75 22 1100 412 710 1419 646
I6 UD S 5 7 23 45 12 12 25 45 90 53
Internal Geometric mean 48% 76% 214% 427% 100% 42% 44% 86% 172% 100%

Please cite this article in press as: Hansen, O.R., et al., Estimation of explosion loading on small and medium sized equipment from CFD
simulations, Journal of Loss Prevention in the Process Industries (2015), http://dx.doi.org/10.1016/j.jlp.2015.12.004
 O.R. Hansen et al. / Journal of Loss Prevention in the Process Industries xxx (2015) 1e17
To quantitatively evaluate the performance of the 4 methods
predicting the loads at 13 external and 6 internal objects for each
simulation, a geometric mean average of the predictions compared
to observations (Method 5-DLM) is reported as follows:

Qn 1
Predicted load n
Geometric Mean ¼ Q1n
1 Observed load
If this geometric mean is 100% this would mean that the average
predicted load level with the method is perfect, even if there may
still be a scatter, with overestimation for some objects and under-
estimation for others. A geometric mean of 50% would tell that the
method would on average underestimate the loads by a factor of
two.
For the strong explosion in the upper table of Table 2 part of the
exercise is also to check the effect of not using the new recom-
mended guidelines for accurate blast wave propagation. The ex-
plosion loads predicted based on the simulation using standard
guidelines are compared against the estimated loads Method 5-
DLM using the FLACS guidelines for improved blast predictions
proposed by (Hansen and Johnson, 2015). For the internal explosion
loads the new proposed guidelines for blast propagation have no
effect at all, as expected, while for the external objects there is an
average underestimation of explosion loads of 14% using the stan-
dard guidelines of FLACS. The worst deviation is for the far field
objects (BP4 48 m East and BP6 43 m NE) where predicted loads
based on DLM increased from 30 and 22 kPa to 54 and 39 kPa which
is a maximum load increase of 80% due to the sharper blast waves.
The underestimation increases with distance, at 24 m the new
guidelines give 35e50% higher pressures. Very few consultants
using FLACS will follow these new proposed guidelines as they have
not yet been endorsed by the model developers, which means that
in most simulation studies one should expect a factor of two
underprediction for reported differential pressure loads at 50 m
distance, and even more at locations further away.
By evaluating the results some very clear trends are seen. For
explosion loads for objects located inside the flame zone a
reasonable estimate of maximum explosion load can be obtained
from the form drag using a drag coefficient Cd ¼ 2.0. Method 2-
Drag underestimates the weak explosion internal maximum loads
by 24% and the strong explosion loads by 22%. The internal loads in
the DDT scenario is underestimated by more than a factor two
(56%), this could be because shock waves from the detonating gas
pocket cause some of the maximum loads. Shockwave loading has a
different character than explosion wind. If Method 1-DfP is used to
estimate drag loads from reported static pressures in the flow field
near the object, the underprediction is twice as much as for Method
2-Drag, 52% for the weak explosion, 42% for the strong explosion
and 58% for the DDT scenario. For the internal explosion loads the
methods based on static pressure give much too high estimated
explosion loading, Method 3-P and Method 4-2P predict more than
2 and 4 times too high explosion loads for the weak explosion
scenario, and almost as much overprediction (1.8 and 3.7 times) for
the strong explosion scenario, while for the DDT scenario the
observed explosion loads are in most cases near what is predicted
by Method 3-P or 4-2P.
For objects located outside the explosion the drag based
methods, Method 1-DfP and Method 2-Drag, give consistently
much too low load predictions. Method 2-Drag gives a factor of
three too low loads (despite using Cd ¼ 2.0) for the DDT scenario, a
factor four for the strong explosion and a factor eight too low
predicted loads for the weak explosion, while Method 1-DfP gives
on average 5 to 6 times too low predictions for all cases. For loading
on objects outside the explosion these approaches are thus very
misleading and should not be used for load estimates. Pressure
Please cite this article in press as: Hansen, O.R., et al., Estimation of explosion loading on small and medium sized equipment from CFD
simulations, Journal of Loss Prevention in the Process Industries (2015), http://dx.doi.org/10.1016/j.jlp.2015.12.004
7
(2)
Fig. 4. Reflection factor in air for shocks with normal incidence towards a surface.
based methods give much more reasonable results for such objects
outside the explosion. For the weak explosion, where pressure
waves will not have a shock front, the Method 3-P seems to give
reasonable load estimates for all external targets with a 12%
underprediction on average, while for the strong explosions the
explosion loading is between that estimated by Method 3-P and
Method 4-2P. For the DDT scenario Method 4-2P gives the best
estimate (geometric mean 101%) with some underprediction for
the highest pressures and overprediction for lower pressures in the
direction opposite of the flame exit where venting, not shock
waves, is the main load contribution. A possible approach for tar-
gets outside the explosion is to use a pressure based method with
reflection factor 1.0 for explosions weaker than 50 kPa at source,
while for stronger explosions where shockwaves can be expected a
reflection factor of two or higher should be used, for instance as
found in Lee's Loss Prevention in the Process Industries (Mannan
and Lees, 2006) for shock waves in air with normal incidence,
Po ¼ overpressure in shock (barg), Pr ¼ reflected overpressure
(barg), Pa ¼ ambient pressure (bara), see also Fig. 4:
ð7pa þ 4po Þ
pr ¼ 2po (3)
ð7pa þ po Þ
From this evaluation we have the following conclusions:
 Drag relations from explosion pressure, as described in FABIG
TN-08 and DNV RP-D101 are not sufficiently conservative for
pairs of reported static pressure and drag, even with a simplified
relation proposed by DNV RP-D101 assuming drag to be at least
1/3 of pressure. In the comparisons shown in this article drag
would frequently be a factor of two or more higher than pre-
dicted from the proposed relation. If the maximum reported
overpressure anywhere on walls/decks inside the explosion is
used to estimate a maximum drag in the area, this can be more
representative, but not always conservative, for maximum
observed drag inside the area.
 For explosion loads inside the flame zone, and possibly outside
the flame at locations in a semi-confined volume (e.g. offshore
module or building) which is pressurized by the explosion, a
relation based on form drag with a conservative drag coefficient
(Cd ¼ 2.0 for box shape, and likely 1.2 for cylinder) may give a
reasonable, but not conservative, estimate of the maximum
explosion load. For the 1 m box-objects evaluated in our study,
estimated loads were 20e25% too low for weak and strong
explosions.
 For locations outside the flame zone where the explosion load
comes from blast waves emitted from the flame zone or build-
ing/module of the explosion, the maximum explosion loading
can be estimated based on the static side-on pressure near the
8 O.R. Hansen et al. / Journal of Loss Prevention in the Process Industries xxx (2015) 1e17

object with a reflection coefficient of two or higher for strong

explosions generating shockwaves. For weak explosions
(<50 kPa at the source) no reflection factor may be required.
 Predicted loads in the far-field from strong explosions gener-
ating shock-waves will be severely underestimated if the stan-
dard FLACS guidelines are applied. In our example this would be
a factor of two 50 m away from the explosion, even more at
larger distances. For such situations the proposed improved
guidelines for far-field blast predictions should be used.

Challenges for the engineer are as follows:

 The suggested relations above give a reasonable estimate for

maximum load, but not the shape and duration. While the
transient form drag load may correlate quite well with the DLM
inside the explosion, it is more challenging to extract the far-
field load dynamics on an object from a static pressure
measurement.
 Another challenge is to judge when form drag can be used and
when the pressure based methods shall be used. Inside a semi-
confined module the form drag relations may be acceptable as
long as the explosion is not too strong, while for a more open
scenario, like an FPSO/FLNG, the pressure based relations should
be used closer to the flame zone. For a study with 100s of sim-
ulations such judgments must be automated.
 For more detailed assessments it will also be a wish to extract
time varying load in a direction or in a plane (e.g. across a pipe).
In the following sections an improved, generalized method to
predict explosion loading on objects is proposed and validated
against the simulation test matrix.
4. An improved and generalized method for load estimation
In the general drag force formula (1) the form drag is normally
the most important term for small equipment inside the flame
zone. With increasing object size term 2, accelerated mass, and
term 4, static pressure differences, start to become more important
as it takes more time for the pressure to equalize around the object.
For far-field pressures, these pressure differences in the flow field
may be the most important contribution, while form drag may be of
secondary importance. This particularly applies for shockwaves
with zero rise time. To formulate a general expression for explosion
loading valid both inside the flame zone and outside in the far-field,
the expression should contain both form drag elements and a
transient pressure term. It is also important that all parameters can
be extracted from nearby monitor points in FLACS. To achieve more
precise load estimates for smaller objects inside or outside the
flame zone, the following general expression is proposed for the
force per area onto an object:
Force/area(t) ¼ Drag(t) þ Ps(t) - Ps(t-Dt)
where Drag is Term 1 from (1), form drag, and Ps is the static free-
field pressure. Dt is the time it takes for a pressure difference to
equalize around an object, and will depend both on the dimensions
of the object, but also on the shockwave velocity, the temperature
dependent speed of sound and the direction of the force.
For a 1 m cube shaped box in the far-field, the distance pressure
must travel from the front of an object to the back is between 1 m
and 2 m, see Fig. 5, and by integration it can be found that the
average shortest travel distance from any point at the front face to
any point at the back is 1.33 m. As the flow will need to round
corners we can assume that the characteristic distance for pressure
to equalize will be slightly higher, possibly of the order 1.4e1.5 m.
Fig. 5. Illustration of distance for pressure to travel from the front to the back on a box
shaped object (left), and definition of dimensions for a non cubical box (right).
For an elongated box profile 1 m  1 m across the average shortest
travel distance is 1.50 m, plus the rounding of path.
A general formula for average shortest path from front to back of
box shaped object can be derived to:
 
Dmin D
Dist ¼ 1
min þ L (5)
2 3Dmax
with Dmin and Dmax being shortest and longest dimension of side
facing the load and L the length of the object along the flow path. To
this minimum distance a constant Kedge is multiplied (~1.1), to take
into account the fact that flow will find slightly longer optimal
paths around the corners.
In order to properly estimate the total force to an object, there is
a need first to estimate directional forces (which can partly be
positive and partly negative) and thereafter combine the forces into
a total force or a force in a plane relevant for loading on pipes and
beams. In this process distances around the object according to (5)
need to be calculated for all three axial directions. For an incoming
pressure/drag load from an arbitrary angle the time for pressure to
equalize around an object will depend on the angle of approach, for
a pressure wave going parallel to a pipe, the time should be zero for
equalizing around the pipe, while for a pressure wave approaching
a 1 m  1 m  1 m box diagonally (45 angle) the wave will start
approaching the back of the cube 0.71 m after approaching the
front, but it will take somewhat longer for the pressure to fill the
entire rear surface than to fill the front, thus, the time for pressure
equalization e.g. in the x-direction can be expected to be between
71% and 100% of the time it would take for a wave hitting the box
head-on, thus as a detailed calculation is challenging the following
assumption is done:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Forcedir
Distdir;angle ¼ signðForcedir Þ Distdir dir ¼ x; y; z (6)
Force
(4) From the variables extracted from monitor points the direc-
tional form drag will be the most suitable for the estimation of the
direction of the force, thus Forcedir/Force is replaced by Dragdir/
Drag.
The purpose of identifying the distance from the front to the
back of an object was to find an estimate for the time for a pressure
difference to equalize, thus in addition to a distance there is also a
need to estimate how fast the pressure will propagate. The sonic
velocity in a mixture primarily consisting of nitrogen, mainly de-
pends on the temperature of the gas, and there is a change of speed
of sound by a factor ~3 at the moment of flame arrival, and shock
waves of different strengths will travel at different speeds. These
aspects need to be included when estimating the time for pressure
to equalize, to estimate the sonic velocity the gas density r is

Please cite this article in press as: Hansen, O.R., et al., Estimation of explosion loading on small and medium sized equipment from CFD
simulations, Journal of Loss Prevention in the Process Industries (2015), http://dx.doi.org/10.1016/j.jlp.2015.12.004
 O.R. Hansen et al. / Journal of Loss Prevention in the Process Industries xxx (2015) 1e17 9

reported at the monitor points for every time step of the simula-
tion, together with pressure and drag components. Also notice that
the Dtdir can both be negative and positive depending on direction
of directional forces/drag, i.e. sign(Forcedir) of equation (6). The
evaluation of the method presented in Chapter 5 indicates that a
fairly good estimate of wave velocity is obtained by using an esti-
mated sonic velocity as follows:
Distdir pffiffiffiffiffiffiffiffiffiffiffi
Dtdir ¼ Kedge dir ¼ x; y; z & Csound ¼ gP=r
Csound ðP; rÞ
The monitor point should be placed so that it is exposed to
almost the same flow field as the object, but at the same time far
enough away from the object so that the presence of the object will
not change the flow field at the monitor point. For the current
evaluation study monitor points would mostly be located adjacent
to the center of the object, between 2 and 3 grid cells away from its
surface. For an early phase explosion study a recommended
approach would be to distribute regular arrays of monitor points
reporting the required variables, in addition to placing points near
objects of interest. Optimally points should be placed in relatively
open areas and not too close to objects strongly influencing the
flow.
5. Evaluation of proposed method for load estimation
(7) In this section explosion loads on objects of various size/shape
are estimated based on free-field pressure, drag (directional com-
ponents) and density r using the proposed load estimation method,
and will be compared to direct load measurements (DLM) reported
in simulations. The entire transient load history is predicted, not
only the maximum load. To evaluate the performance of the
method for different flow/loading regimes a number of different
loading targets will be evaluated both for the weak explosion, the
strong explosion (using new proposed guidelines for accurate far-
field blast) and the DDT scenario, as described earlier in this article.

Fig. 6. Predicted loads on external objects B1 (6 m East, upper), B3 (24 m East, middle) and B5 (24 m NE, lower) from pressure, drag and density (solid black line) compared to direct
load measurements (black dotted line). Static (red) and dynamic (blue) pressure at nearby monitor points used to estimate the load are also shown. Weak explosion (left), strong
explosion (middle) and DDT-scenario (right) are shown. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Please cite this article in press as: Hansen, O.R., et al., Estimation of explosion loading on small and medium sized equipment from CFD
simulations, Journal of Loss Prevention in the Process Industries (2015), http://dx.doi.org/10.1016/j.jlp.2015.12.004
10 O.R. Hansen et al. / Journal of Loss Prevention in the Process Industries xxx (2015) 1e17

5.1. Total loads on 1 m  1 m  1 m cubical objects outside the
explosion
In Figs. 6 and 7 load estimates (black solid lines) for external
cubical objects B1, B3, B5, B8, W1 and W2 (all at ground level) based
on reported static pressure (blue dotted line) and drag (red dotted
line) at a nearby monitor point. This is compared to the reported
direct load measurements (black dotted line) based on panel
pressures reported on its four walls in x- and y-direction. Vertical
direction is not included in the estimate for these objects as the
objects are located on the ground. The comparisons are performed
for weak explosions (400010) with maximum pressure around
50 kPa, strong explosion with maximum pressure around 250 kPa
and a DDT scenario with maximum pressures reported around
3 MPa. By studying the plots in Fig. 6 it can be seen that the tran-
sient load estimates (black solid lines) in general represent very
good predictions of the reported transient loads (black dotted
lines), in particular for B3 (24 m East) and B8 (12 m North), while
for B1 (6 m East) which is just outside the gas cloud and will
experience a strong venting of combustion products an error of
20e25% in maximum load can be seen. For the DDT case a load of
3.1 MPa (31 bar) is reported at the object, while our method pre-
dicts only 2.2 MPa. One possible explanation to this is that the
detonating pocket of gas extends 3e5 m outside the rig, while the
front of object B1 starts 5.5 m from the rig and our monitor point to
report pressure and drag is located above B1 6 m from the rig. In the
first couple of meters outside a detonating gas cloud, there may be a
significant pressure reduction within 0.50 m, and thus, much of the
deviation could be explained by the fact that the monitor point
used for the load estimate is 0.50 m further away from the deto-
nating cloud than the front of B1 and 2.25 m above the ground. For
B3 and in particular B5 it can be seen that multiple peaks are
reproduced very well. Load estimates based on pressure or drag
alone would not give meaningful results for these. Looking at the
loads from the strong explosion at target B3 or B5, it can be seen
that the duration of pressure and drag loads is around 30e40 ms,
while the duration of the loads at the objects (reported and esti-
mated) is around 10 ms.

Fig. 7. Predicted loads on external objects B8 (12 m North, upper), W1 (10 m South near reflecting wall, middle) and W2 (15 m SE near reflecting wall, lower) from pressure, drag
and density (solid black line) compared to direct load measurements (black dotted line). Static (red) and dynamic (blue) pressure at nearby monitor points used to estimate the load
are also shown. Weak explosion (left), strong explosion (middle) and DDT-scenario (right) are shown. (For interpretation of the references to color in this figure legend, the reader is
referred to the web version of this article.)

Please cite this article in press as: Hansen, O.R., et al., Estimation of explosion loading on small and medium sized equipment from CFD
simulations, Journal of Loss Prevention in the Process Industries (2015), http://dx.doi.org/10.1016/j.jlp.2015.12.004
 O.R. Hansen et al. / Journal of Loss Prevention in the Process Industries xxx (2015) 1e17 11

In Fig. 7 similar plots are shown for B8 (12 m North of rig), W1
and W2 located South/South-East close to a 4 m tall wall in the
terrain, so that multiple reflections will be expected to give a
chaotic load pattern. For the weak and strong explosions the pre-
dicted transient loads are very similar to the reported direct loads,
and studying the static and dynamic pressures it seems very chal-
lenging to predict such a loading pattern from these. For the DDT
scenario two out of three maximum loads are within a 30% error,
but the transient load pattern is a little less good. One main
explanation for this is likely that the direction of the strong
shockwaves from the detonating gas pocket will arrive from a
different direction than the flow direction at the time, and thus that
the Dt used to estimate the added difference in static pressure is not
accurate.
5.2. Total loads on 1 m  1 m  1 m cubical objects inside the
explosion
I5 in the SW corner (lower deck) and I6 towards the South (upper
deck) are shown in Fig. 8. All these objects are located away from
solid surfaces, and the load estimate is therefore based on all three
axial directions. Studying the plots it should become clear that the
pressure (red dotted line) is not a particular useful parameter for
the prediction of internal explosion loads, and even the drag (blue
dotted line) using a conservative drag coefficient, is far from suffi-
cient to predict the details of the transient loads. In most cases our
proposed method for load estimate gives a good prediction of
maximum load and transient development, some underprediction
is however seen for I3 and I6 regarding shock waves from the
detonating gas pocket, and there is also a second peak of the load
on I6 which is not well picked up, the reason for this could be that
the monitor point is at a location not fully exposed to the flow/
pressure pattern which leads to that second peak.
5.3. Loads on objects with non-regular shape outside the explosion

The next group of loading objects are internal targets, and to So far only regular cubical boxes are considered. Among the
show the performance of the prediction methodology for these, targets defined were also a number of less regular shapes, e.g. T1
predicted and reported loads for I3 NNE in the test rig (upper deck), and T3 which were 3 m long rectangular pipes (0.5 m diameter)

Fig. 8. Predicted loads on internal objects I3 (NNE-UD, upper), I5 (SE-LD, middle) and I6 (S-UD, lower) from pressure, drag and density (solid black line) compared to direct load
measurements (black dotted line). Static (red) and dynamic (blue) pressure at nearby monitor points used to estimate the load are also shown. Weak explosion (left), strong
explosion (middle) and DDT-scenario (right) are shown. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Please cite this article in press as: Hansen, O.R., et al., Estimation of explosion loading on small and medium sized equipment from CFD
simulations, Journal of Loss Prevention in the Process Industries (2015), http://dx.doi.org/10.1016/j.jlp.2015.12.004
12 O.R. Hansen et al. / Journal of Loss Prevention in the Process Industries xxx (2015) 1e17

Fig. 9. Predicted loads on 3 m  0.5 m  0.5 m rectangular pipes outside explosion, T1 (along blast direction) and T3 (across blast direction) 13 m East of rig (solid black line)
compared to direct load measurements (black dotted line). Static (red) and dynamic (blue) pressure at nearby monitor points used to estimate the load are also shown. Strong
explosion (left) and DDT-scenario (right) are shown. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

located along (T1) and across (T3) the blast propagation direction
13 m East of the test rig. In Fig. 9 plots are shown for the strong
explosion (left) and the DDT scenario (right) showing that for the
strong explosion the force estimates for the rectangular pipes are
very good, and that the force is roughly four times higher if the pipe
is located across the flow than along. The cross-sectional area is 6
times higher, thus the accelerated mass force contribution seems to
have an impact for the pipe along the blast propagation direction.
For the DDT case there is some overestimation of the force on the
pipe, in particular for the case with the pipe along the blast
direction.
In Fig. 10 similar plots are shown for two 2 m  2 m  0.5 m
plates located 33 m East of the test rig along and across the blast
propagation direction. Both for the strong explosion and the DDT
the load estimates are within roughly 20% of the reported with a
reasonable duration of load. As the deviation seems to be some-
what higher for the non-regular objects compared to the cubical
objects, there may be a need for more work to optimize the for-
mulas for distance for the pressure to equalize.
5.4. Scaling effects, 1 m cube and 2 m cube compared
There was also a 2 m  2 m  2 m cube among the external
targets, located around 18 m away from the test rig (slightly to-
wards NE). In Fig. 11 the loading on this is compared to the smaller
1 m  1 m  1 m cube located at 24 m distance. From the plots it
can be seen that for all four cases (strong explosion and DDT sce-
nario, small and large cube) the explosion loads are well predicted.
It can be seen that the duration of the load is 30e50% longer for the
larger object. For the larger object the plots of drag and static
pressure have been multiplied by four (front surface area of cube)
to make a comparison to the total force more meaningful.
5.5. Evaluation of directional load estimates
As already mentioned all the loads presented have been calcu-
lated by first estimating the directional loads and thereafter
combining these. In order to evaluate how well directional loads
are estimated some examples are shown in Fig. 12 in which x-di-
rection (red), y-direction (blue) and total (black) estimated loads
(solid) are compared to directly reported (dotted) loads based on
differential pressures across the objects (directional or combined).
For this evaluation both weak explosions, strong explosions and the
DDT scenario are included, and targets located at an angle to the
axis of the explosion are evaluated (B5 24 m NE, B6 43 m NE, W1
10 m SSE and W2 15 m SE). The latter two objects are close to a
reflecting wall and therefore challenging to properly predict.
By studying the plots it can be seen that most loads, directional
as well as combined, are well predicted, in particular for objects B5
and B6. For the more challenging W1 and W2 object

Please cite this article in press as: Hansen, O.R., et al., Estimation of explosion loading on small and medium sized equipment from CFD
simulations, Journal of Loss Prevention in the Process Industries (2015), http://dx.doi.org/10.1016/j.jlp.2015.12.004
 O.R. Hansen et al. / Journal of Loss Prevention in the Process Industries xxx (2015) 1e17 13

Fig. 10. Predicted loads on 2 m  2 m  0.5 m plates outside the explosion, T2 (along blast direction) and T4 (across blast direction) 33 m East of rig (solid black line) compared to
direct load measurements (black dotted line). Static (red) and dynamic (blue) pressure at nearby monitor points used to estimate the load are also shown. Strong explosion (left) and
DDT-scenario (right) are shown. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

loads more scatter is seen, however, maximum loads seem well explosion, and four times higher for the DDT scenario.
predicted. It should be observed that the directional loads can both For the internal targets x- and y-loads are also well predicted,
be negative and positive, while the total load is always positive. with geometric mean of 104% and 86% for the weak explosion, 111%
and 85% for the strong explosion, and 71% and 117% for the DDT
scenario. A main reason for the underprediction in the x-direction
5.6. Summary of all targets
for the DDT-scenario is that the strong blast waves sent back in the
negative x-direction from the detonating gas pocket seems not
In Tables 3e5 tabulated values from the weak explosion, strong
properly picket up. For the total loads the geometric means (esti-
explosion and DDT scenario are presented, both the reported drag
mated/observed) are 94%, 111% and 65%, the lower value for the
and pressure at monitor points, and approximate duration, direc-
DDT-scenario likely caused by strong shockwaves from the deto-
tional load estimates compared to reported loads, and finally a total
nation strongly underpredicted at target I4. One possible cause for
load estimate compared to reported total load. For objects located
this is that the monitor point used for the load estimate at target I4
at ground level only horizontal forces are considered.
was placed in the shadow behind object I4 and protected from the
It can be seen that estimates of external loads are very good,
shockwaves from the gas pocket.
geometric mean (estimated/observed) for directional loads in x-
To summarize the performance of the proposed method it must
and y-direction is 97% and 113% for the weak explosion, 103% and
be concluded that the load predictions based on drag, pressure and
105% for the strong explosion and 89% and 110% for the DDT-
density reported from a nearby monitor point seem surprisingly
scenario. The total load estimates for external targets are 102%,
accurate for a range of different loading cases, both the maximum
98% and 86% respectively, with the largest deviation for the DDT
value, the transient time history with multiple peaks, as well as
scenario. For reported loads above 20 kN (kPa) the geometric mean
directional components. It should be expected that many of the
of the estimated load duration versus the reported load duration
estimated loads would be less accurate as the monitor point could
have been compared, which was 108% (4 objects) for the weak
receive a different flow pattern or be shielded from shockwaves,
explosion, 103% (13 objects) for the strong explosion, and 99% for
compared to what is received at the target, in particular for internal
the detonation case. For comparison the duration of the drag and
targets.
pressure pulses at the monitor point is around two times higher
A feasibility study to evaluate the lack of accuracy and need for
than the force duration for the weak explosion and the strong

Please cite this article in press as: Hansen, O.R., et al., Estimation of explosion loading on small and medium sized equipment from CFD
simulations, Journal of Loss Prevention in the Process Industries (2015), http://dx.doi.org/10.1016/j.jlp.2015.12.004
14 O.R. Hansen et al. / Journal of Loss Prevention in the Process Industries xxx (2015) 1e17

Fig. 11. Predicted loads on 2 m  2 m  2 m cube 18 m outside explosion (T5) and 1 m  1 m  1 m cube 24 m outside explosion (solid lines) compared to direct load reporting
(dotted black lines). Static pressure (red dotted) and dynamic pressures (blue dotted) are scaled up with front area of the box. Strong explosion (upper plots) and DDT scenario
(lower plots). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

improvements of current best-practice approaches has been per-
formed. The principles of a new, improved approach to estimate
explosion loading on pipes and small equipment has been estab-
lished and an initial evaluation of method has been performed.
Further work is planned to improve and optimize the current
approach, one element to be improved is the formulas for load
equalizing for non-regular objects. So far only rectangular objects
are considered, the aim will be to find similar relations for cylin-
ders, grated decks, beams and pipe bundles.
6. Summary and conclusions
Work has been initiated to assess current practices for the
estimation of explosion loading onto pipes and smaller equipment.
One part of the activity is also to develop improved methods esti-
mating loads from CFD explosion studies. Current findings include:
 Current industry guidelines for estimation of loads on equip-
ment not properly resolved in a CFD-simulation (e.g. FABIG TN-
08) are challenging to apply in a consistent and conservative
way. A generalized drag formula is referred to, but only form
drag is thereafter considered, even if other contributions can
also be important.
 A formula to estimate drag as function of overpressure is pro-
posed in these guidelines, but this is generally very non-
conservative if applied for individual locations inside and
outside an explosion. Only if the maximum pressures at walls,
decks or corners are used to estimate maximum drag values,
reasonable estimates can be found with this relation.
 Loads inside an explosion can often be reasonably well esti-
mated by the use of reported dynamic pressure (form drag)
using a conservative drag coefficient (this gave 20e25% under-
prediction in our evaluation). Loads in the far field can mostly be
estimated using (side-on) monitor point pressures with an
appropriate reflection coefficient, but the load duration may be
strongly overestimated. If loads are predicted using form drag
outside the flame ball for low confined platform modules or
FPSOs, which is often done in explosion studies, a strong un-
derestimation of loads can be expected.
 Far-field loads reported from strong explosions with FLACS us-
ing standard guidelines used by most consultants, will be
severely underpredicted due to smearing of blast waves. With
new proposed guidelines reducing the numerical time steps
once the blast waves leave the flame ball, twice as high
maximum loads can be expected at 50 m distance.
 An improved method for load prediction has been proposed and
tested against simulations in which drag, pressure and density
from monitoring points are used to estimate loads on object
surfaces. The method is tested on box objects inside and outside

Please cite this article in press as: Hansen, O.R., et al., Estimation of explosion loading on small and medium sized equipment from CFD
simulations, Journal of Loss Prevention in the Process Industries (2015), http://dx.doi.org/10.1016/j.jlp.2015.12.004
 O.R. Hansen et al. / Journal of Loss Prevention in the Process Industries xxx (2015) 1e17 15

Fig. 12. Predicted directional loads on external objects B5 (24 m NE, upper), B6 (43 m NE, 2nd row), W1 (10 m SSE, 3rd row) and W2 (15 m SE, lower). Black curves are total loads,
red curves x-direction and blue curves y-direction. Estimated results are shown using solid curves, while reported direct load measurements are shown as dotted lines. Weak
explosion (left), strong explosion (middle) and DDT-scenario (right) are shown. (For interpretation of the references to color in this figure legend, the reader is referred to the web
version of this article.)

the explosion with good results, both for maximum and direc-
tional load and duration.
 The concept of the new proposed methodology for explosion
loading is to distribute numerous monitor points in the domain
to report transient flow parameters, from these the transient
loading on actual or future objects of any location, type, size and
orientation can be estimated with good precision.
 A further optimization of the proposed methods is initiated, for
instance for non-cubical shaped boxes, and further work is
required to identify how to handle cylindrical objects, and loads
onto grated decks, pipe bundles and other groups of objects.
Detailed CFD blast studies have been carried out for more than
20 years for the offshore oil and gas industry, however the gap

Please cite this article in press as: Hansen, O.R., et al., Estimation of explosion loading on small and medium sized equipment from CFD
simulations, Journal of Loss Prevention in the Process Industries (2015), http://dx.doi.org/10.1016/j.jlp.2015.12.004
16 O.R. Hansen et al. / Journal of Loss Prevention in the Process Industries xxx (2015) 1e17

Table 3
Weak explosion, maximum monitor point values, reported and estimated directional and total loads, on targets as well as approximate load durations for all external (upper)
and internal (lower) targets. The geometric mean averages are calculated for estimated vs observed loads and duration, as well as monitor point drag, pressure and their
duration versus reported differential pressure load.

Weak explosion (400010) Reported from monitor point (kPa) Directional load kN (¼kPa except T1- Total load kN (¼kPa except T1-T5)
T5)

Object Shape Position Drag (ms) Pressure (ms) DLM Estimated DLM Estimated DLM (ms) Estimated (ms)

Max Duration Max Duration DP-X DP-X DP-Y DP-Y DP-total Duration DP-total Duration

BP1 1 m cube 6mE 11.6 80 32 110 18 23 2 2 19 40 24 55

BP2 1 m cube 12 m E 4.4 70 23 110 10 10 3 1 10 60 10 60
BP3 1 m cube 24 m E 1.3 50 14 100 5 4 1 1 5 4
BP4 1 m cube 48 m E 0.3 50 13 70 3 3 1 3 3 5
BP5 1 m cube 24 m NE 0.7 60 10 110 3 2 2 2 4 3
BP6 1 m cube 43 m NE 1.3 70 10 80 2 2 1 1 2 3
BP7 1 m cube 6mN 4.6 150 12 150 6 8 6 11 6 11
BP8 1 m cube 12 m N 1.2 150 8 150 1 2 3 3 3 3
BP9 1 m cube 24 m N 0.2 150 6 150 2 1 2 1 2 1
BP10 1 m cube 48 m N 0.1 150 6 150 1 1 2 1 2 1
BP11 1 m cube 25 m NW 0.1 150 5 150 1 1 1 1 1 1
R1 1 m cube 6mW 0.9 320 8 200 2 3 1 1 2 2
R2 1 m cube 12 m W 1.0 220 6 200 2 4 1 0 2 3
W1 1 m cube 10 m SSE 1.6 180 12 110 4 3 6 5 6 7
W2 1 m cube 15 m SE 1.3 120 13 90 5 5 3 3 6 6
T1 3  0.5  0.5 m 13 m E 2.9 110 20 130 3 3 3 3 4 4
T2 0.5  2  2 m 33 m E 0.6 110 10 90 5 4 3 3 5 5
T3 0.5  3  0.5 m 13 m E 3.4 80 20 130 9 10 1 2 9 10
T4 2  2  0.5 m 33 m E 0.6 70 10 90 12 9 1 2 12 15 9 15
T5 2 m cube 18 m ENE 1.4 70 14 80 31 25 12 18 32 60 35 60
External Geometric Mean 27% 153% 289% 225% DP-X 97% DP-Y 113% DP-max 102% DP-duration 108%
I1 1m cube LD NW 0.4 200 10 200 2 1 1 1 2 1
I2 1 m cube UD NW 3.4 180 10 200 4 7 1 2 5 7
I3 1 m cube UD NE 8.8 180 31 110 15 17 12 13 20 150 21 150
I4 1 m cube UD SE 9.0 180 39 110 16 23 10 9 20 150 25 150
I5 1 m cube LD SE 8.6 170 38 110 20 19 10 8 22 130 20 130
I6 1 m cube UD S 3.7 180 23 140 11 10 6 2 12 250 10 200
Internal Geometric Mean 38% 108% 213% 71% DP-X 104% DP-Y 86% DP-max 94% DP-duration 95%

Table 4
Strong explosion, maximum monitor point values, reported and estimated directional and total loads, on targets as well as approximate load durations for all external (upper)
and internal (lower) targets. The geometric mean averages are calculated for estimated vs observed loads and duration, as well as monitor point drag, pressure and their
duration versus reported differential pressure load.

Strong explosion (400101) Reported from monitor point (kPa) Directional load kN (¼kPa except T1- Total load kN (¼kPa except T1-T5)
T5)

Object Shape Position Drag (ms) Pressure (ms) DLM Estimated DLM Estimated DLM (ms) Estimated (ms)

Max Duration Max Duration DP-X DP-X DP-Y DP-Y DP-total Duration DP-total Duration

BP1 1 m cube 6mE 89 25 132 50 173 212 8 7 173 25 214 25

BP2 1 m cube 12 m E 44 25 102 40 126 137 10 9 126 20 138 25
BP3 1 m cube 24 m E 19 25 71 30 115 99 6 10 115 8 100 8
BP4 1 m cube 48 m E 4 40 33 40 49 37 11 10 49 8 37 8
BP5 1 m cube 24 m NE 8 40 45 35 40 38 31 29 50 15 48 15
BP6 1 m cube 43 m NE 3 35 27 40 25 22 22 20 33 15 30 15
BP7 1 m cube 6mN 11 160 25 120 13 21 14 19 16 22
BP8 1 m cube 12 m N 4 110 17 100 7 8 13 11 14 11
BP9 1 m cube 24 m N 1 90 12 120 2 2 7 4 7 5
BP10 1 m cube 48 m N 1 140 10 70 1 2 5 3 5 4
BP11 1 m cube 25 m NW 1 90 6 120 4 4 2 2 4 4
R1 1 m cube 6mW 2 250 13 160 15 17 3 1 6 5
R2 1 m cube 12 m W 3 200 12 160 30 33 2 2 5 8
W1 1 m cube 10 m SSE 8 80 43 60 5 8 20 15 24 40 23 40
W2 1 m cube 15 m SE 13 60 56 45 50 29 19 12 35 55 34 55
T1 3  0.5  0.5 m 13 m E 28 35 82 35 25 31 17 18 32 50 36 50
T2 0.5  2  2 m 33 m E 11 30 52 35 77 70 21 73 79 10 98 12
T3 0.5  3  0.5 m 13 m E 35 35 88 35 125 128 5 14 128 40 133 45
T4 2  2  0.5 m 33 m E 9 30 52 35 328 250 14 41 329 8 259 7
T5 2 m cube 18 m ENE 16 35 62 35 391 325 137 202 416 40 425 40
External Geometric Mean 26% 201% 125% 221% DP-X 103% DP-Y 105% DP-max 98% Duration 103%
I1 1 m cube LD NW 1 230 17 200 4 3 3 3 5 4
I2 1 m cube UD NW 7 130 18 200 8 14 3 4 10 15
I3 1 m cube UD NE 30 100 90 70 56 67 33 35 68 105 76 110
I4 1 m cube UD SE 32 100 122 60 66 94 41 30 72 85 100 90
I5 1 m cube LD SE 35 100 131 60 74 93 38 35 83 65 93 65
I6 1 m cube UD S 9 130 45 60 25 22 13 6 27 120 24 120
Internal Geometric Mean 40% 117% 181% 68% DP-X 111% DP-Y 85% DP-max 111% Duration 103%

Please cite this article in press as: Hansen, O.R., et al., Estimation of explosion loading on small and medium sized equipment from CFD
simulations, Journal of Loss Prevention in the Process Industries (2015), http://dx.doi.org/10.1016/j.jlp.2015.12.004
 O.R. Hansen et al. / Journal of Loss Prevention in the Process Industries xxx (2015) 1e17 17

Table 5
DDT scenario, maximum monitor point values, reported and estimated directional and total loads, on targets as well as approximate load durations for all external (upper) and
internal (lower) targets. The geometric mean averages are calculated for estimated vs observed loads and duration, as well as monitor point drag, pressure and their duration
versus reported differential pressure load.

DDT scenario (400200) Reported from monitor point (kPa) Directional load kN (¼kPa except T1- Total load kN (¼kPa except T1-T5)
T5)

Object Shape Position Drag (ms) Pressure (ms) DLM Estimated DLM Estimated DLM (ms) Estimated (ms)

Max Duration Max Duration DP-X DP-X DP-Y DP-Y DP-total Duration DP-total Duration

BP1 1 m cube 6mE 722 10 721 15 3041 2083 121 223 3042 4 2210 4
BP2 1 m cube 12 m E 215 14 275 15 579 686 43 72 579 5 695 5
BP3 1 m cube 24 m E 54 22 123 20 259 227 41 42 259 6 230 6
BP4 1 m cube 48 m E 6 40 40 40 64 49 7 11 64 8 50 8
BP5 1 m cube 24 m NE 35 25 106 20 161 155 132 140 208 6 215 6
BP6 1 m cube 43 m NE 6 40 43 30 58 50 51 47 77 7 69 7
BP7 1 m cube 6mN 46 110 54 110 73 84 50 48 84 25 87 25
BP8 1 m cube 12 m N 30 80 57 100 58 25 200 94 208 15 95 15
BP9 1 m cube 24 m N 10 50 41 90 10 10 49 42 50 25 43 25
BP10 1 m cube 48 m N 3 110 21 80 5 6 28 21 28 8 21 8
BP11 1 m cube 25 m NW 5 70 22 120 30 17 14 13 32 25 19 25
R1 1 m cube 6mW 8 180 23 180 15 16 6 3 15 25 16 25
R2 1 m cube 12 m W 5 160 25 180 31 11 6 4 31 20 11 25
W1 1 m cube 10 m SSE 67 65 153 60 55 145 178 120 187 18 153 18
W2 1 m cube 15 m SE 79 30 162 35 292 261 240 98 378 20 265 20
T1 3  0.5  0.5 m 13 m E 178 17 232 20 114 144 68 161 121 30 228 25
T2 0.5  2  2 m 33 m E 21 19 74 25 125 114 62 125 126 9 164 8
T3 0.5  3  0.5 m 13 m E 194 17 266 15 746 938 18 52 747 5 945 5
T4 2  2  0.5 m 33 m E 17 19 73 25 537 406 27 75 538 6 422 5
T5 2 m cube 18 m ENE 55 18 123 20 1167 883 385 601 1224 8 1197 9
External Geometric Mean 20% 413% 58% 444% DP-X 89% DP-Y 110% DP-max 86% Duration 99%
I1 1 m cube LD NW 3 200 38 150 22 12 8 11 23 20 15 20
I2 1 m cube UD NW 11 250 28 150 22 20 13 12 25 150 21 150
I3 1 m cube UD NE 81 80 120 70 200 117 85 126 209 90 146 90
I4 1 m cube UD SE 80 60 452 40 782 226 188 182 1086 22 297 25
I5 1 m cube LD SE 206 60 710 40 252 563 275 527 646 8 845 5
I6 1 m cube UD S 13 160 45 150 46 30 26 19 53 120 30 120
Internal Geometric Mean 22% 272% 86% 201% DP-X 71% DP-Y 117% DP-max 65% Duration 94%

between small equipment drag loads and large object pressure
differences has never been satisfactorily handled in these studies.
This article demonstrates how these two types of loads can be
merged into one relation (4) so that accurate loads for medium
sized equipment can be obtained. It is believed that the findings
presented can have an important impact on the way engineering
companies, ship-yards and oil companies will choose design
strength for installations, and that the methods, when properly
developed, will represent significant improvements to current best
practices.
References
API-RP-2FB, 2006. Recommended Practice for the Design of Offshore Facilities
against Fire and Blast Loading. The American Petroleum Institute.
API-RP 752, 2009. Management of Hazards Associated with Location of Process
Plant Permanent Buildings. The American Petroleum Institute.
Baker, W.E., Cox, P.A., Westine, P.S., Kulesz, J.J., Strehlow, R.A., 1983. Explosion
Hazards and Evaluation. Elsevier Scientific Publishing Company.
Czujko, J. (Ed.), 2001. Design of Offshore Facilities to Resist Gas Explosion Hazard e
Engineering Handbook.
DNV RP-D101, 2008. Structural Analysis of Piping Systems. Det Norske Veritas,
October 2008.
DNV OS-A101, 2014. Safety Principles and Arrangements. Det Norske Veritas, July
2014.
FABIG TN-08, Protection of Piping Systems Subject to Fires and Explosions, The Steel
Construction Institute, Silwood Park, Ascot, Berkshire, SL5 7QN, United
Kingdom.
Hansen, O.R., Talberg, O., Bakke, J.R., 1999. CFD-based methodology for quantitative
gas explosion risk assessment in congested process areas: examples and vali-
dation status. In: International Conference and Workshop on Modelling the
Consequences of Accidental Releases of Hazardous Materials. 28 September-1
October 1999. AIChE, CCPS, San Francisco, California, pp. 457e477.
Hansen, O.R., Gavelli, F., Davis, S.G., Middha, P., 2013. Equivalent cloud methods
used for explosion risk and consequence studies. J. Loss Prev. Process Ind. 26
(2013), 511e527.
Hansen, O.R., Johnson, D.M., 2015. Improved far-field blast predictions from fast
deflagrations, DDTs and detonations of vapor clouds using FLACS CFD. J. Loss
Prev. Process Ind. 35, 293e316.
Interim Guidance Notes for the Design and Protection of Topside Structures against
Fires and Explosions, The Steel Construction Institute, Publication Number SCI-
P-112, 1992.
ISO 19901-3, 2010. Petroleum and Natural Gas Industries - Specific Requirements
for Offshore Structures - Part 3: Topsides Structure. The Int. Org. for
Standardization.
Mannan, Sam, Lees, Frank P., 2005. Lee's Loss Prevention in the Process Industries.
Elsevier Butterworth-Heinemann, Burlington, MA.
NORSOK Z-013, 2010. Risk and Emergency Preparedness Analysis. Norsok standard.
Available from: Standard Norge, Postboks 242, N-1326 Lysaker, Norway.
Sand, I.Ø. (1999). Received loading test cases. Fire and Blast Information Group
(FABIG) Technical Newsletter, Issue 24, R347: 14e18.
Yasseri, S. (2002), Response of Secondary Systems to Explosion, FABIG Newsletter
Article No. R454, FABIG Newsletter Issue No. 33, September 2002.

Please cite this article in press as: Hansen, O.R., et al., Estimation of explosion loading on small and medium sized equipment from CFD
simulations, Journal of Loss Prevention in the Process Industries (2015), http://dx.doi.org/10.1016/j.jlp.2015.12.004