1st CCPS APCPS ______________________________________________________________________

A Consequence Based Assessment Method/Software Model to

Facilitate the Safe Location and Design of Occupied Buildings on or
In Close Proximity to Process Facilities

Kehinde Shaba
Det Norske Veritas Software
Palace House, London, UK
kehinde.shaba@dnv.com

Colin Hickey
Det Norske Veritas Software
Palace House, London, UK
colin.hickey@dnv.com

[Insert here any copyright claims you wish to assert]

Prepared for Presentation at

1st CCPS Asia-Pacific Conference on Process Safety
Kempinski Hotel, Qingdao, China
September 4 – 5, 2013

UNPUBLISHED

AIChE shall not be responsible for statements or opinions contained

in papers or printed in its publications
1st CCPS APCPS ______________________________________________________________________

A Consequence Based Assessment Method/Software Model to

Facilitate the Safe Location and Design of Occupied Buildings on or
In Close Proximity to Process Facilities

Kehinde Shaba
Det Norske Veritas Software
Palace House, London, UK
kehinde.shaba@dnv.com

Colin Hickey
Det Norske Veritas Software
Palace House, London, UK
colin.hickey@dnv.com

Keywords: Facility Siting, VCE modeling, Multi-Energy model, Baker Strehlow model,
Consequence-based analysis methods, Occupied Building Assessment

Abstract
Ensuring the safe location and design of occupied buildings (permanent or temporary) is high on
the agenda of key stakeholders – industry, regulators, and the workforce, not to mention
members of the public who have a vested interest in the area. The significant amount of
analytical work required in support of such an analysis represents a key barrier to conducting
such studies in a detailed manner. It is often the case that a range of simplifications are utilised
that can result in a less than realistic assessment.

This paper describes the implementation of two of the most widely used vapour cloud explosion
(VCE) models, namely the TNO multi-energy (ME) and Baker-Strehlow-Tang (BST) into Phast,
a general purpose onshore consequence analysis software package developed by DNV Software.
This in direct response to the aforementioned challenges and the intention is to provide a robust
platform by which a more detailed analysis of the VCE threats of process plant can be evaluated
- without recourse for simplification which can be detrimental from a risk management
perspective. The implementations of both models are quite detailed and include all parameters
that have a bearing on the magnitude of a vapor cloud explosion. A simple case study of the tool
as applied to a hypothetical occupied building siting challenge is presented and the value of
using the tool discussed.

Note: Do not add page numbers. Do not refer to page numbers when referencing different portions of the
paper
1st CCPS APCPS ______________________________________________________________________

1. Introduction

Major accidents such as Texas city etc. have focussed attention on the need to ensure that
occupied buildings are well protected from the often devastating impacts of vapour cloud
explosions (VCE’s) arising from process facilities. Ensuring the safe location and design of
occupied buildings (permanent or temporary) is high on the agenda of key stakeholders –
industry, regulators, and the workforce, not to mention members of the public who have a vested
interest in the area.

In recent times, much has been done to develop and advance the knowledge gaps in this area.
Notable projects include GAME (Eggen, 1988), GAMES (Mercx, et al., 1998), RIGOS (Van den
Berg & Mos, 2005) to name a few. Yet, a key issue that remains is that the degree of analytical
effort required in support of a detailed building evaluation can be significant due to the extensive
time and cost required. This is largely due to the often complex, multi-system, large nature of
process plant – there is “lots going on” and thus “lots to consider”. It is clearly an unintended
consequence of the prevailing process plant design ethos. Adding to the complexity is the
intricate nature of the VCE accident sequence. A single loss of containment event can result in
numerous outcomes – VCE or otherwise depending on factors such as the prevailing weather
conditions, wind direction, degree of interaction with one or more regions of congestion or
confinement and the location of ignition sources. Where a VCE outcome is realized, the possible
impact to each building of interest has to be evaluated. It is often the case that most of the factors
identified above are present in multiples e.g. a range of scenarios, weather conditions, obstructed
regions, buildings of concern etc. The wide range of factors that need to be considered (“in
isolation” or “as an interaction with other factors”) also has a significant bearing on the level of
effort required.

Various techniques are employed to make this task more manageable. The first is “screening
exercises”. Whilst there is much to be gained from making effective use of limited resources,
screening is generally most useful and less harmful for screening hazards that clearly do not
present VCE potential either intrinsically (e.g. degree of flammability) or by virtue of their
context (no presence of a congested or confined region) - but can be dangerous when used
beyond this point (i.e. in grey areas) largely because it becomes more subjective and thus less
precise – whose standards of conservatism are being applied? How conservative is conservative
enough? To be used effectively it also requires high levels of expertise and competence.

Another involves “scenario rationalisation”. Much emphasis tends to be placed on establishing a

representative set of scenarios that can be considered to be sufficiently reflective of the whole.
Again, whilst this is useful, there is a danger that the selected scenarios might not be as
representative as thought, leading to an underestimation of the threat level posed.

A third involves using “simplified models” which in some cases are then subject to further
simplification. Much of the tediousness is made worse by the fact that the underlying physics of
explosions is complex and this is reflected in the equally complex nature of the predictive
models which are available. Computational fluid dynamic based models are generally considered
to give the best approximations of this complex phenomenon. However, such models are largely
considered impractical for day to day work due to the significant resource and time requirements
1st CCPS APCPS ______________________________________________________________________

needed and are generally only justifiable for highly uncertain issues. Additionally, the resource
demands of this route is such that applying it for a plant wide assessment is generally not
feasible. Simplified models are designed to solve this problem by allowing a robust level of
characterization without the added expense. But then again there are varying degrees of
simplicity and even the simplified models tend to require a wide range of inputs. The result is
further simplification - it is common to place emphasis on the key parameters that are thought
have a significant influence. An example is the implementation of the TNO multi energy model
into the existing base configurations of Phast (versions 7.01 and below) which utilises the size of
the confinement and blast curve number as the sole, key governing parameters. Additionally
such models do not often examine the interaction of the flammable cloud with the region of
confinement or confinement. Furthermore, such implementations tend to be conservative in their
outlook and focus solely on the worst case outcome, a basis not always suitable for risk
management. Together, these factors act as impediments to a realistic assessment of VCE
potential to occupied buildings.

This paper describes a more detailed implementation of two of the most widely used vapour
cloud explosion models, namely the TNO multi-energy (ME) (Van den Berg, 1985) and Baker-
Strehlow-Tang (BST) (Baker, et al., 1996) into Phast, a general purpose onshore consequence
analysis software package developed by DNV Software. This in direct response to the
aforementioned challenges and the intention is to provide a robust platform by which full and
detailed analysis of the VCE threats of process plant can be evaluated - without recourse for
simplification which can be detrimental from a risk management perspective.

2. Background on Explosions

2.1 General

Prior to elaborating on the various methods by which the effects of a vapour cloud explosion can
be estimated, it is useful to examine the nature of explosions in general.

An explosion is essentially a very rapid release of energy that results in a pressure discontinuity
in the atmosphere. This discontinuity—called a “pressure wave”—initiates in the blast source
and dissipates outwards. The energy associated with the wave is the primary source of
destructive energy to objects in its path and is responsible for a large proportion of the damage
that occurs from explosions on process facilities. Consequently, it is important to be able to
characterize the nature of this wave as function of distance.

A pressure wave is typically characterized by three key components. First, the magnitude of the
pressure it generates above ambient referred to as the “overpressure”; the maximum value is
known as the “peak overpressure”. Second, the duration of the wave, simply called the “blast
duration”. Third, the blast “impulse” which is defined as the change in momentum and is the
force-time product. Together these are often referred to as “blast parameters” or “blast load” in
the context of a particular building or location.
1st CCPS APCPS ______________________________________________________________________

There are many types of explosions. Vapour cloud explosions tend to be the most significant for
the bulk of process facilities. In some circumstances, explosions due to dust particles or pressure
vessels can dominate.

2.2 Vapour Cloud Explosions

A wealth of knowledge has been accumulated on this area, such that there is an extensive range
of material that elaborates to a great level of detail, the particulars of the science underpinning
vapour cloud explosions and consequently how they can be modelled. Key references include
(CCPS, 1994), (Lees, 1996) and (Crowl, 2003).The goal of the text that follows is to highlight
the key aspects so as to provide sufficient contextual information to allow the uninitiated reader
develop a better appreciation of the state of the knowledge without recourse to primary sources.

Various factors have a bearing on the potential for VCE’s. Necessary preconditions include: a
flammable vapour cloud (for example following a loss of containment event and sufficient
mixing in air to produce a combustible mixture) and an ignition source. Other factors – some
intrinsic to the material and others related to the environment also play a key role in determining
the magnitude of the resulting overpressures produced. The two key external factors that
influence the magnitude of a VCE are often referred to as the “degree of congestion” and the
“degree of confinement” though it is not uncommon to find them described using disparate
terminology.

A VCE occurs as a result of a flame propagating through a fuel-air mixture. Research into this
area (TNO, 1997) has indicated that the magnitude of the resulting overpressure is directly
related to the flame speed which is governed by the degree of turbulence realized. Turbulence is
generated by the interaction of the flame with obstacles in its path. This interaction involves a
vicious feedback cycle – the more obstacles faced, the higher the degree of turbulence and the
increased acceleration of the flame front. With no turbulence, the cloud will burn as a flash fire
without appreciable overpressures generated. High turbulence results in increased flame speeds
and high overpressures. The opposite occurs with low turbulence. The level of obstacles within
the cloud (i.e. obstacle density) is referred to as the “degree of congestion”, whilst the location
wherein the obstacles are situated is termed a “congested region” or “region of congestion”.

As the cloud burns, it expands. This expansion allows for the rapid dissipation of energy. A
vapour cloud allowed to expand in all 3 dimensions will lead to lower overpressures than
confined ones (say beneath a platform or pipe-rack).The phrases “degree of confinement” or
“degree of expansion” relates to the extent to which the clouds expansion is “unrestricted” or on
the contrary “confined”. Confinement is considered in one or all three dimensions (x, y, z), with
consideration of intermediate positions. It is often the case that there are many opportunities for
such areas on process plants. The number of congested/confined areas identified increases the
calculation effort significantly.

Other factors such as material reactivity and not least the degree of “blanketing” between the
flammable cloud and the region of confinement/congestion also have an influence.
1st CCPS APCPS ______________________________________________________________________

2.3 Key Vapour Cloud Explosion Models

The most widely used simplified models to characterize—that is estimate the magnitude and
duration of the resulting pressure wave as a function of distance from the explosion source—the
nature of vapour cloud explosions are the TNO multi-energy (hereafter referred to as ME) (Van
den Berg, 1985), the Baker-Strehlow-Tang (hereafter referred to as BST) (Baker, et al., 1996)
and the Congestion Assessment Method (CAM) (Puttock, 1995). A fuller description/elaboration
of each of the above models can be found in (CCPS, 1994), (Lees, 1996) and (Crowl, 2003).

The TNT equivalence model, also a simple model and once used extensively has now been
shown to be limited in its ability to adequately characterize VCE’s. Consequently, it is not
generally recommended for use. For example API standard 752 (API, 2009) explicitly states that
the model should not be used. Lees (1996) asserts that the “TNT equivalent method for vapour
cloud explosion is obsolescent”. Nevertheless, it still sees widespread application. This is in part
driven by its simplicity – only a few inputs are required to the model. That it results in relatively
conservative estimates of the peak overpressure (often used as a key indicator of explosion
magnitude) is another factor – one that makes it especially appealing as an early screening tool.
The sections that follow briefly outline the key premise and characteristics of the models that
have been integrated into the Phast consequence package in a more robust manner, namely the
ME and the BST models.

2.3.1 TNO multi-energy model

The TNO multi-energy model (Van Den Berg, 1985) is based on the supposition that the energy
of an explosion is governed largely by the level of confinement and congestion (i.e. location
specific factors) within which the vapour cloud is located and to a lesser extent by the intrinsic
characteristics of the vapour cloud, most notably the flammable mass. By placing more emphasis
on external factors, it represents a significant shift in thinking from the notions animating the
TNT model.

The Multi Energy explosion model requires the identification of the obstructed region, volume
blockage ratio, degree of expansion among other parameters as inputs. These are then utilised to
calculate the overpressure generated with the level of congestion (input) provided. Alternatively,
the model can use the “curve number” as input. The “Curve number” refers to different
overpressure versus distance graphs. These graphs differ in range of severity from “Curve 1-
weak explosion” to “Curve 10-severe/heavy explosion”. Materials that may be released will be
assigned a specific “curve number”, i.e. strength of the explosion, depending on their properties
such as their reactivity.

2.3.2 Baker Strehlow Tang model

The BST model (Baker, et al., 1996) also allows for the characterization of overpressure as a
function of distance but uses the flame speed as the point of departure, the selection of which is
based on three factors. First, the reactivity of the material of concern. Second, the potential for
flame expansion within the region of the release (i.e. degree of confinement) and third, the
obstacle density (degree of congestion).
1st CCPS APCPS ______________________________________________________________________

In this model, the potential for expansion is characterized by the number of dimensions in which
expansion is possible and includes one of four possibilities that sit on a spectrum from most
confined (known as “1D”, which reflects tunnels, corridors or similar geometries where
expansion is only possible in one direction) to completely open (“3D” which reflects a situation
where there is minimal restriction to expansion and the cloud can expand in all directions).
Occupying the middle area is the “2D” and “2.5D” designations which aim to capture
intermediate positions between these two extremes.

For the most part the ME and BST model are quite analogous, particularly in terms of their
outlook and animating concepts. The ME model is known to give higher predictions of
overpressure in the near field. The BST is also known to over predict, albeit in the far field.

3. The Phast Software Package and Recent Updates

3.1 The Software package “Phast”

Phast (acronym for Process Hazard Analysis Software Tool) is a consequence analysis software
package developed by DNV Software for the modelling of the consequences associated with
releases of flammable or toxic materials as would occur on process facilities.

Phast's modelling capabilities are numerous and diverse and allow for the evaluation of a wide
range of hazard types in different situations. Its capabilities cover a wide range of release phases
(gas, liquid and two-phase) and materials (various flammables and toxic compounds including
LNG and CO2) as pure components or in mixtures. In terms of release types, it can model
unpressurised and pressurised releases as steady state or as time-dependent releases.

At its core is the Unified Dispersion Model (UDM) which enables complete and rigorous
modelling of various release types accounting for transition to jet, dense and passive clouds;
buoyancy effects, substrate interaction (ground effects); plume lift-off, capping at the
mixing/inversion layer, droplet formation and rainout under diverse atmospheric conditions.

Phast is an integrated consequence analysis package that models all stages of an accidental
release from:
• Discharge: including rainout, pool evaporation and spreading to
• Dispersion: from the pool and the discharge orifice as shown in Figure 1.
• Effects including toxicity, jet fires, fireballs, pool fires, flash fires and explosions
1st CCPS APCPS ______________________________________________________________________

Figure 1: Initial stages of an accidental release as modelled by Phast

In terms of VCE modelling, Phast enables the key requirement namely the size and dimensions
of the dispersing vapour cloud to be established.

The modelling takes into account various contextual factors that can influence the development
of the hazard. More details of the technical capabilities of the Phast Software package can be
found in body of technical reference documentation available as well as in papers by Witlox
(2010) and Witlox & Oke (2008) amongst others.

3.2 Recent Updates

As indicated in the introductory section, the current base configurations of Phast (versions 7.01
and below) include what can effectively be referred to as a narrow implementation of both the
ME and BST models. In the former, the blast curve number and the volume of confinement are
considered to be key governing parameters to the exclusion of others such as degree of
congestion. For the latter, the flame speed and confined volume are given primacy. Also, in both
implementations, the interaction between the dispersion cloud and the region of
congestion/confinement—a key element as the quantity of flammable mass located therein is a
governing factor—is quite opaque and defined in a rather ambiguous manner. Additionally, the
epicentre of the explosions was not located around the region of congestion but rather within the
1st CCPS APCPS ______________________________________________________________________

cloud – quite contrary to what would be expected in reality where the explosion centre will be
located within the area of confinement. Finally, whilst the tool had the capability to evaluate
numerous scenarios simultaneously and display the hazard range as contours on a location map,
evaluating the impact to any one target was generally a tedious affair as this has to be done on an
individual case-by-case basis – a herculean task where a large number of scenarios are being
analysed.

The software package has now been updated to address the limitations highlighted above. Key
innovations with regards to VCE modelling include: a fully featured and exhaustive
implementation of the ME and BST explosion models as well as the inclusion “multi-scenario
contouring capability”. These updates are available as an add-on to the Phast base configuration
for version 7.1 onwards. The implementation of both models into Phast is executed as part of the
Obstructed Region Explosion Model (OREM) developed by DNV Software. The theory and
validation of the OREM model is discussed in (Xu, 2010a) and (Xu, 2010b) respectively.

The validation work involved comparison of the results derived with the OREM model with
those from seminal experimental studies such as BFEST3a, EMERGE and the Shell Deer Park
Ethylene case (as reported in (Fitzgerald, 2001)), and the GAMES project (as reported by
(Mercx, et al., 1998)). The GAMES study included verification versus three sub cases: a
chemical plant case, a gas processing case and the LNG terminal case. Overall, the model
showed good agreement with the experimental data in terms of its ability to predict the blast
parameters. Better agreement was observed for the ME model than for the BST model.

Key aspects of these models are the definition of obstructed regions, consideration of the degree
of overlap between the flammable cloud and the region so as to determine the explosive potential
as well as consideration of critical separation distance between defined regions (features
common to both models). The separation distance is important as it aids in the determination of
how many explosion sources can be reasonably expected where a vapour cloud envelopes more
than one region of obstruction. Finally, given the fact that the model will lend itself to the
consideration of a wide range of scenarios, the software has been updated to allow for multi-
scenario contouring - an efficient way of organizing and communicating the resulting hazard
ranges especially when used to inform facility siting considerations as outlined in CIA (2010)
(American Petroleum Institute (API), 2009) and (American Petroleum Institute (API), 2007)).
All of these are discussed in the sections that follow.

3.2.1 Definition of “Obstructed Regions”

As noted above, “obstructed regions” (i.e. regions of congestion and/or confinement) are central
to the potential for high strength VCE’s and are thus a key element of both the ME and BST
models. They are typically defined based on a review of the site layout diagrams/plot plans and
the external boundaries of these areas are typically defined via “perimeter bounding boxes”.

Whilst there are guidelines to facilitate this activity (see for example (Van den Berg & Mos,
2005) and (American Petroleum Institute (API), 2007), there remains a degree of subjectivity
which can lead to some variation in the definition depending on the perspective of the person
1st CCPS APCPS ______________________________________________________________________

undertaking the task. Once defined, they are then approximated to distinct cuboid shaped
volumes and input to the Phast model using the geographical information system (GIS).

Beyond the geometrical similarities in obstructed regions for both explosion models they both
require a different set of parameters for their complete definition. As a consequence they are
created differently within the Phast software package. Additionally, for each explosion model,
two types of obstructed regions can be created: “Defined Strength” or “Calculated Strength” in
the case of the ME model and “Defined Flame Speed” or “Calculated Flame Speed” in the case
of the BST model. For the TNO multi energy model, key parameters for the “Calculated
Strength” option include degree of expansion (or confinement), flame path length and volume
blockage ratio (together used to determine the curve number that will apply). For the “Defined
strength” option, the only inputs required are the curve number and the blockage ratio. For the
BST model, the “Calculated Flame Speed” option requires degree of confinement, congestion
level and flame reactivity are key (used to calculate the flame speed that will obtain). The flame
speed and the blockage ratio are the only inputs required “Defined Flame Speed” option.

To sum, the following logic applies within Phast:

1) Obstructed regions are created as either ME or BST
2) For both model types, two types of obstructed regions can be created – “Defined” or
“Calculated”.
a. For ME, these are referred to as “Defined Strength” or “Calculated Strength”
b. For BST, these are called “Defined Flame Speed” or “Calculated Flame Speed”

A simple hypothetical case study has been developed to highlight how the tool will work from a
practical standpoint. The objective of the study is to evaluate the side-on overpressure loads to
three existing buildings of interest arising from a number of release events associated with a
proposed installation of a gas compression facility. The loads are to be used as basis for
consideration of possible mitigation strategies.

A range of pressurized ethane release scenarios associated with the proposed facility has been
identified. These are located in and around the congested regions. For each scenario, three hole
sizes have been considered – Small, Medium and Large. A single weather condition has been
assumed. In terms of the wind rose, 8 directions have been accounted for.

Figure 2 illustrates the four obstructed regions (shown as pink boxes) defined for this study. The
figure also shows the three occupied buildings of interest to this study.
1st CCPS APCPS ______________________________________________________________________

Figure 2: Graphical User Interface (GUI) representation of obstructed regions in Phast*.

*Annotations highlighting the positions of obstructed regions and buildings of interest have been added to improve clarity

3.2.2 Consideration of flammable cloud/obstructed region overlap

It has been established that overlap of a flammable cloud with a region of confinement or
congestion is a necessary prerequisite for a VCE of appreciable overpressure to occur,
consequently it is important to evaluate the extent to which such blanketing occurs for the range
of scenarios to be considered and obstructed regions defined on a facility.

A range of factors have an influence on this possibility. The relative distance between the release
event and the obstructed region, the size of the resulting flammable cloud as it develops and
disperses and the direction of travel of the cloud are key aspects to consider. The range of
possibilities for a single dispersing flammable cloud travelling downwind with a single
obstruction and building in its path is illustrated in Figure 3 for four representative time steps
(i.e. snap shots) over the duration of the cloud. The situation becomes more complex, where
multiple events, multiple obstructed regions and multiple wind directions are considered.
1st CCPS APCPS ______________________________________________________________________

Figure 3: Possible cloud steps/outcomes as a flammable cloud disperses

Phast incorporates a sophisticated GIS system into which release locations and obstructed
regions are defined. The system enables the model to examine in detail the degree of interaction
between the resulting flammable clouds and the obstructed regions and thus determine the degree
of fill to be used within the explosion calculations. It does so for all wind directions specified
(four as a minimum; 18 as a maximum).

3.2.3 Critical Separation Distance

Under the right conditions, each identified obstructed region has the potential to act as an
explosion source. However, where a dispersing flammable cloud overlaps more than one
obstructed region there is the prospect for multiple explosion sources. The potential is governed
by the “separation distance” between the regions. A large gap between multiple regions will
mean that the flame front will decelerate resulting in multiple independent explosions. On the
other hand, a small gap will result in a single explosion with the combined energy of all the sub-
sources. The “critical separation distance” is the threshold criterion used to establish whether an
explosion source should be considered as independent or part of a larger group.

There are divergent views on the specification of the critical separation distance. TNO (1997)
suggests a critical separation distance equal to 10 obstacle diameters with an upper bound of
25m. This follows on from their definition of a congested region as one in which obstacles are
1st CCPS APCPS ______________________________________________________________________

located within 10 obstacles diameters of one another. Subsequent research into this area
(specifically the work titled “Research into Guidance on Separation Distances, the “RIGOS”
project, see Van den Berg et al, 2002) has indicated that this approach might not always be
conservative particularly where the resulting overpressures are high. The report recommends the
critical separation distance should be a function of the size of the maximum overpressure from
the source explosion, referred to as a “critical separation ratio”. Baker et al (1998) suggests than
5m should be adopted as the upper limit for defining obstructed regions are separate explosion
sources.

The Phast model allows for the critical separation distance to be specified as a “ratio” or “fixed
value”. In the former case, users have to determine the ratio to apply based on a preliminary
assessment of the maximum overpressures created within each region.

Based on the above, the following additional logic applies to how obstructed regions can be
defined in the software tool:
1) Obstructed regions can be defined as being independent or as part of a group.
2) Where defined as part of a group, the critical separation criterion has to be specified to
determine the number of potential sources within the group.
3) The criterion can be defined as affixed distance or a ratio.
4) If the criterion is not specified, all regions within that group are considered to be single
source
5) Only obstructed regions defined within the same group are considered for combination
potential. Regions in different groups or sets cannot be combined.
6) To avoid “double counting” resulting in possible overestimation of the explosion volume,
overlap between obstructed regions within a set is not allowed. Violations are
automatically checked within the software.

3.2.4 Multi-Scenario Hazard Contouring

Another key update is the capability to plot combined hazard contours to a defined threat level
(e.g. as overpressure or impulse) for a range of scenarios on a facility layout diagram. This
feature allows for clearer understanding and communication of hazards ranges and their impact
footprint as well as makes for more efficient and comprehensive analysis - the full range of
flammable release scenarios (as opposed to a limited subset) can be modelled simultaneously.

Table 1 below indicates the pressure calculation methods available to output the blast parameters
(overpressure and impulse). This feature is not limited to explosion hazards only but can also be
used to develop aggregate hazard contours for dispersing clouds (e.g. flammable or toxic) to a
concentration of interest and thermal radiation from fires (jet and pool fires).

The side-on overpressure contours derived for the hypothetical study referred to in section 3.2.1
above is shown in Figure 4. Plotting the contours on a location map readily enables the
overpressure loads buildings can potentially experience to be identified. The result for this
hypothetical study indicates:
1st CCPS APCPS ______________________________________________________________________

 The peak overpressure potentially experienced by the three buildings of interest will be
of the order of 0.3 bar (this contour just marginally touches the periphery of these
buildings.).
 The 0.2 bar overpressure contour overlaps most of the buildings suggesting that this is a
more representative pressure loading.
 Due to the fact that the 0.3 bar contour just skirts the borders of the buildings, some
further investigation of the results will be required to evaluate how sensitive this
overpressure level is to key assumptions (e.g. size/degree of congestion of the obstructed
region).
 Additionally, it is evident the blast loads to these buildings are driven by the obstructed
congestion 1 (see Figure 2). Reducing the potential for congestion/confinement in this
area represents a potential solution to mitigating the blast loads to the buildings of
interest.

Table 1: Output options for blast parameters Overpressure and Impulse for ME and BST
modelling options
Pressure Method  ME  BST 
Side on Overpressure  Yes  Yes 
Dynamic Overpressure  Yes  No 
Side on + Dynamic Overpressure  Yes  No 
Reflected Overpressure  Yes  Yes 

A robust evaluation of explosion hazard potential to buildings requires a consideration of both

the “overpressure level experienced as well as the duration” or the “overpressure and duration”.
Though not shown, the software model includes the capability to plot impulse contours for the
range of overpressure methods available above.
1st CCPS APCPS ______________________________________________________________________

Figure 4: Side-on overpressure contour envelope for multiple scenarios and weather directions
(representative result)*

*Annotations highlighting the positions of buildings of interest have been added to improve clarity.

4. A new method to facilitate the “consequence-based” evaluation of VCE

blast parameters to occupied buildings

Taken together, the above innovations allow for a new approach to the evaluation of blast loads
to occupied buildings using the Phast Software package – one that is not necessarily limited by
the often large number of events and obstructed regions that obtain on process facilities. Using
this software model, the following simple three steps are required:

 Step 1: Establish calculation basis. Identify all release scenarios as well as areas of
congestion/confinement. Select explosion model to be used (i.e. ME or BST). Input into
the model together with parameters (e.g. separation distance between obstructed regions)
required for their full definition.
 Step 2: Undertake modelling. Phast models each event in turn and examines the VCE
potential and where found, determines the associated hazard ranges and blast loads as a
function of downwind distance.
1st CCPS APCPS ______________________________________________________________________

 Step 3: View results and determine blast loads to buildings. The results are presented
on the location maps as contours for blast parameters of interest (e.g. Overpressure and
Impulse).

It is good practice to examine the sensitivity of results to key input parameters/ assumptions,
consequently, some iteration between steps 1 and 3 is also recommended.

The overall consequence of the above is that a more detailed evaluation of VCE blast hazards to
all buildings on a facility resulting from all loss of containment events postulated can be realised
simultaneously using this tool. This is particularly noteworthy given that such a task often
represents a key challenge requiring significant amount of resources (cost and time).

Recognising that Phast is a consequence analysis tool, this advancement will facilitate the
execution of building siting analyses that are predicated upon “consequence-based” evaluation
methods. Other methods include “Simplified methods” (e.g. lookup tables that codify industry
experience) and Risk-based methods (full consideration of both the likelihood and impact of an
event e.g. using exceedance curves/hazard frequency contours). Safeti (Cavanagh, 2001;
Worthington and Witlox, 2002), a risk assessment and consequence analysis package already
incorporates this capability and is available for use in risk-based methods. The implementation of
these VCE models into Safeti (versions 6.6 and above) is discussed by Cavanagh et al (2009).

5. Conclusion
Major accidents such as Texas city etc. have focussed attention on the need to ensure that
occupied buildings are well protected from the often devastating impacts of vapour cloud
explosions (VCE’s) arising from process facilities. The large scale of such installations often
implies that a substantial amount of analytical effort is required. Consequently, there is a need to
make the effort much more manageable. This is usually achieved via some kind of screening
activity or the use of much simpler approaches (e.g. using in-house or other standards). Whilst
there is much to be gained from simplifying the analysis, there is a danger that if not executed
properly, a less than adequate assessment of the hazard potential will be achieved. This is
because such screening requires judgment and expertise (“good engineering judgement”) and
such skills are not widely available. Hence, it is desirable to have a software tool which allows
for a substantive analysis to be undertaken in a rigorous, robust and resource (cost, time)
efficient manner without the added risk simplification can introduce.

This paper has described and examined the recent advances in the Phast consequence model
designed to address this challenge. It has demonstrated that increased rigour can be achieved in
an effective manner for any scale study (both large and small) thus negating the need for
screening/simplification activities. The advancements to the Phast software platform are a
welcome addition to the range of software tools available to facilitate the evaluation of explosion
risks to buildings and should help improve the robustness and overall quality of such evaluations
resulting ultimately in more effective management of risks to occupied buildings.
1st CCPS APCPS ______________________________________________________________________

6. References
1) American Petroleum Institute (API), 2007. Management of Hazards Associated with
Location of Process Plant Portable Buildings, API Recommended Practice RP753, 1st
Ed (Reaffirmed January 2012), s.l.: American Petroleum Institute .
2) American Petroleum Institute (API), 2009. Management of Hazards Associated with
Location of Process Plant Buildings, API Recommended Practice RP752, 3rd Ed, s.l.:
American Petroleum Institute.
3) Baker, Q., Doolittle, C., Fitzgerald, G. & Tang, M., 1998. Recent development in the
Baker-Strehlow VCE analysis methodology. Process Safety Progress, pp. 17(4), 297-
301.
4) Baker, Q., Tang, M., Scheier, E. & Silva, G., 1996. Vapor cloud explosion analysis.
Process Safety Progress, pp. 15(2), 106-109.
5) Cavanagh, N. J., 2001. Calculating Risks. Hydrocarbon Engineering (6).
6) Cavanagh, N., Xu, Y. & Worthington, D., 2009. A Software Model For Assessing
Fatality Risk From Explosion Hazards Using The Multi Energy Method And Baker
Strehlow Tang Approach. s.l., IChemE, pp. 647-656.
7) Center for Chemical Process Safety (CCPS) , 1994. Guidelines for Evaluating the
Characteristics of Vapor Cloud Explosions, Flash Fires and BLEVEs. New York: CCPS
AICHE.
8) Center for Chemical Process Safety (CCPS), 1996. Guidelines for Evaluating Process
Plant Buildings for External Explosions and Fires. New York: CCPS AICHE.
9) Chemical Industries Association (CIA), 2010. Guidance for the location and design of
occupied buildings on chemical manufacturing sites 3rd Edition, London: Chemical
Industries Association.
10) Crowl, D. A., 2003. Understanding Explosions. New York: CCPS AICHE.
11) Eggen, J., 1988. GAME: development of guidance for the application of the multi-energy
method (TNO Report PML 1995-C44 Prepared for the UK Health and Safety Executive:
Research Report 202/1998), Rijswijk: TNO Prins Maurits Laboratory.
12) Fitzgerald, G., 2001. A Comparison of Simple Vapor Cloud Explosion Model Explosion
and Prediction methodologies. Texas, s.n.
13) Lees, F. P., 1996. Loss Prevention in the Process Industries: Hazard Identification,
Assessment and Control. Vol. 2 (2nd ed). Oxford: Butterworth-Heinemann Ltd.
14) Mercx, W., Van den berg, A. & Van Leeuwen, D., 1998. Application of correlations to
quantify the source strength of vapour cloud explosions in realistic situations. Final
report for the project: ‘GAMES’. TNO report PML 1998-C53, Rijswijk: TNO Prins
Maurits Laboratory.
15) Puttock, J., 1995. Fuel Gas Explosion Guidelines - the Congestion Assessment Method.
Manchester, s.n.
16) Tang, M. & Baker, Q., 1999. A New Set of Blast Curves for Vapour Cloud Explosions.
Process Safety Progress, Issue 18, pp. 235-240.
17) TNO (The Netherlands Organization of Applied Scientific Research), 1997. Methods for
the calculation of Physical effects due to releases of hazardous materials "Yellow Book"
CPR 14E, The Hague: s.n.
18) Van den Berg, A., 1985. The multi-energy method: a framework for vapor cloud
explosion blast prediction. Journal of Hazardous Materials, pp. (12) 1-10.
1st CCPS APCPS ______________________________________________________________________

19) Van den Berg, A. & Mos, A., 2005. Research to improve guidance on separation
distance for the multi-energy method (RIGOS) (TNO Report PML 2002-C50 Prepared
for the UK Health and Safety Executive: Research Report 369), Rijswijk: TNO Prins
Maurits Laboratory.
20) Witlox, H. W., 2010. Overview of consequence modelling in the hazard assessment
package Phast. s.l., s.n.
21) Witlox, H. W. & Oke, A., 2008. Verification and validation of consequence models for
accidental releases of hazardous chemicals to the atmosphere. s.l., s.n.
22) Worthington, D. & Witlox, H. W., 2002. SAFETI Risk Modelling Documentation –
Impact of Toxic and Flammable Effects, London: DNV Software.
23) Xu, Y., 2010a. Obstructed Region Explosion Model (OREM): Theory, London: DNV
Software.
24) Xu, Y., 2010b. Obstructed Region Explosion Model (OREM): Validation, London: DNV
Software.
25) Xu, Y., Worthington, D. & Oke, A., 2009. Correcting The Predictions By Baker-
Strehlow-Tang (Bst) Model. s.l., IChemE, pp. 318-325.