Microsoft Word - 1.docx
CHEMICAL ENGINEERING TRANSACTIONS
VOL. 77, 2019
A publication of
The Italian Association
of Chemical Engineering
Online at www.cetjournal.it
Guest Editors: Genserik Reniers, Bruno Fabiano
Copyright © 2019, AIDIC Servizi S.r.l.
ISBN 978-88-95608-74-7; ISSN 2283-9216
Vapour Cloud Explosions – The Evidence for Deflagration to
Detonation Transition
Michael Johnsona*, Andrzej Pekalskib, Vincent H.Y. Tamc, Bassam A. Burgand, Pol
Hoorelbekee, Chris Savvidesf
a DNV GL Spadeadam Testing and Research, Brampton, Cumbria, CA8 7AU, UK
b Shell Research Ltd., London, UK
c University of Warwick, University Road, Coventry, CV4 7AL, UK
d Steel Construction Institute, Silwood Park, Buckhurst Road, Ascot SL5 7QN, UK
e Total, Paris la Défense cedex, France
f BP Exploration, Chertsey Road, Sunbury-on-Thames TW16 7LN, UK
michael.johnson@dnvgl.com
Approximately forty years ago, it was realised that the generation of high flame speeds would result in
damaging pressures being produced in vapour cloud explosions (VCEs). At the time, however, it was not clear
how high flame speeds could be generated in VCE. Research in the 1970s and 1980s showed that if a vapour
cloud engulfed a region of congested process pipework, the flame could accelerate to levels where damaging
pressures were produced. Since that time, the accepted basis for assessing VCE hazards has been to
consider congested process regions as potential locations of deflagrations, with pressure reducing as the
explosion propagated away from the region. However, this was not sufficient to explain the evidence found
following the Buncefield explosion in the UK (2005). The research conducted following Buncefield, the most
extensive for any VCE, led to the conclusion that the incident involved a transition from a deflagration to a
detonation (DDT). In reaching this conclusion, several possibilities were considered and dismissed, including a
speculative explosion mechanism not previously observed that involved ignition by radiation of dust particles
ahead of the flame. Taken in combination with more recent experimental studies, they indicate that DDT was
not just an explanation for Buncefield, but likely has a much wider relevance. The evidence obtained from
Buncefield and subsequent research supporting DDT will be reviewed.
1. Introduction
That damaging overpressures can be generated following the ignition of a flammable cloud in a confined
volume has been understood for some time. However, in the second half of the 20th century, it became clear
that large vapour cloud explosions in more open environments were not fully understood.
It was known that high flame speeds could generate damaging overpressure (Kuhl, 1981), however the
mechanism for producing these high flame speeds was not. Laboratory scale experiments at that time
suggested flame speeds of only a few metres a second would be produced in the open, between one and two
orders of magnitude less than was considered necessary to give damaging overpressures.
Subsequent experimental research, from laboratory (Abdel-Gayed, 1982) to large scale (Eyre 1987, Harris
1989), showed the importance of flame distortion and turbulence in flame acceleration. In a vapour cloud that
engulfed unconfined congestion (e.g. process pipework), the flow produced by the expanding flame interacted
with the congestion, resulting in flame distortion and turbulence. This intensified the combustion and
accelerated the flame, potentially to high flame speeds and leading to significant overpressure.
In 2005, a VCE occurred at the Hertfordshire Oil Storage Ltd (HOSL) depot at Buncefield, UK. The HOSL site
had little process congestion and had not been considered to have a VCE hazard. Given the severity of the
explosion and the subsequent multiple tank fires, an independent investigation was carried out by the
Buncefield Major Incident Investigation Board (BMIIB 2006). This investigation was wide ranging but included
DOI: 10.3303/CET1977117
Paper Received: 23 November 2018; Revised: 2 April 2019; Accepted: 8 July 2019
Please cite this article as: Johnson M., Pekalski A., Tam V., Burgan B., Hoorelbeke P., Savvides C., 2019, Vapour Cloud Explosions – The
Evidence for Deflagration to Detonation Transition, Chemical Engineering Transactions, 77, 697-702 DOI:10.3303/CET1977117
697
an initial review the evidence related to the VCE by a group of explosion experts, which identified that the
incident did not appear to be consistent with the standard view that VCEs are restricted to regions of process
congestion (BMIIB 2007).
This prompted a more detailed assessment in a Joint Industry Project (JIP) (HSE 2009) followed by large
scale experimental studies (Burgan 2014). In relation to the explosion, the JIP had three main objectives:
• To determine if it was possible to explain the evidence with a ‘deflagration only’ scenario.
• To determine if DDT was consistent with both the evidence from the incident and the results from
experimental studies.
• To assess other potential combustion mechanisms, particularly a speculative explosion mechanism
not previously observed that involved ignition by radiation of dust particles ahead of the flame, a so
called ‘episodic explosion’.
Supported by experimental studies, it was concluded that the evidence from the incident was not consistent
with the ‘deflagration only’ scenario but was consistent with a deflagration leading to a DDT. No evidence was
found in experimental studies that supported the possibility of the episodic explosion mechanism.
In addition, the JIP research provided extensive data on the overpressure damage to items in and around a
detonating cloud (Burgan 2014) giving a solid relationship between damage and strength of incident pressure
waves. Such understanding, combined with developments in fundamental knowledge of detonation science
has prompted re-investigation of previous intensive vapour cloud explosions accidents (Chamberlain et al,
2018). The damage evidence observed is consistent with the deflagrative and detonative combustion
processes, indicating that most, if not all, large VCEs have involved DDT.
1.1 The problem
Recent UK HSE publications (Atkinson et al 2017a, 2017b) have suggested that neither deflagration in its
‘standard’ form nor DDT could explain Buncefield and some other VCE incidents. They conclude that a new
explosion mechanism (the episodic mechanism mentioned above) was needed. In addition, another
publication has proposed an alternative multi-stage explosion mechanism (Thomas, 2018).
This confusing message has the potential to misdirect the efforts of both researchers and operators of facilities
in proper identification of locations where there is the potential for a severe VCE. The evidence for DDT is
reviewed, showing the high level of consistency that has been found between the research and the
observations from incidents. Comment is also provided on the recent contrary publications.
2. Review of Buncefield
2.1 Why not just deflagration?
The evidence from two incidents is used to illustrate why deflagration alone does not explain some VCEs. As
already identified, the Buncefield incident provides a wealth of evidence. In addition, reference is made to
evidence from a VCE at the Indian Oil Corporation’s (IOC) Petroleum Oil Lubricants (POL) Terminal at
Sanganer in Jaipur, India, which occurred on 29th October 2009 (Johnson, 2012). The evidence that led to the
conclusion that deflagration, on its own, was insufficient to explain these incidents was the presence of severe
overpressure damage and directional indicators throughout the vapour clouds.
Both the Buncefield and Jaipur incidents showed high levels of overpressure damage throughout the vapour
cloud, including damage to vehicles, instrument boxes and oil drums, as illustrated in Figure 1.
Figure 1: Examples of pressure damage at Buncefield (left and centre) and Jaipur (right)
698
The experimental studies demonstrated that the damage could only be caused by overpressures well in
excess of 2 bar. For example, the car damage observed inside the cloud at Buncefield required pressures in
excess of 5 bar. In addition, the damage level to a car within a detonated propane-air cloud closely matched
the damage observed at Buncefield and Jaipur.
Importantly, all the items at Buncefield and Jaipur were in open areas, well away from any congestion that
might sustain a deflagration. Even if a very intense deflagration were postulated in those areas where there
was congestion, the decay in the magnitude of the pressure wave would result in incident pressures on the
items that would have been well below that required to cause the observed damage.
Directional indicators include such things as bent posts, scoured paintwork, fallen trees, trees scorched on
one side and displacement of items laterally. Examples found within the vapour clouds are shown in Figure 2.
Figure 2: Examples of directional indicators from Buncefield (left and centre) and Jaipur
Prior to the involvement of the BMIIB expert group, the directional indicators were misinterpreted. For
example, it was thought that the lamppost post bent over in Figure 2 indicated an explosion travelling from
right to left. This led to the suggestion of at least three separate explosions in an early BMIIB report. However,
modelling conducted by the JIP showed that fast deflagrations and detonations in shallow pancake type
clouds imposed a net negative load on items in the cloud due to the long duration reverse flows behind the
flame caused by expansion of the combustion products. With this reversal, the directional indicators all pointed
to approximately the same location, confirming a consistent, single event explanation.
The indicators were widely spread through the vapour clouds at Buncefield and Jaipur and almost always in
open areas without any congestion. This evidence is not consistent with a standard deflagration scenario
where directional indicators would only be present in the congested regions.
The BMIIB expert group, and subsequently the JIP, therefore considered several alternatives for the VCE at
Buncefield, including DDT, the episodic explosion mechanism and wind generated by buoyancy of fireball.
The last of these was dismissed relatively quickly in the JIP as the inward velocities were insufficient to
produce the deformation caused to the directional indicators. Additionally, any directional evidence would tend
to point towards the centre of the cloud where the buoyancy would be greatest. However, in both Buncefield
and Jaipur, the directional indicators pointed toward the corner or close to the edge of the cloud.
Attention in the JIP was focussed on deflagration leading to DDT, however, in case the DDT scenario turned
out to be inconsistent with the evidence, the speculative episodic mechanism was also studied experimentally.
2.2 DDT
The DDT hypotheses was assessed against a wide range of evidence (Burgan 2014), including:
• The potential for DDT – experimental studies and review of previous experiments (Harris, 1989).
• Directional indicators within the cloud – assessed in both simulations and experiments.
• Severe damage throughout the vapour cloud – detonation experiments assessing damage.
• Timing of events – constraints on timing from CCTV records and witness statements.
• Duration of positive and negative pressure phases – as determined from Buncefield CCTV records.
• Overpressure Decay from the edge of the cloud – assessed using numerical simulation.
Given the consistency of the DDT scenario with the wide range of evidence, the JIP concluded that the DDT
scenario provided the explanation for Buncefield. A similar conclusion was reached for Jaipur.
699
It is worth noting that since the JIP was completed, other experimental programs (Pekalski 2015, Davis 2017)
have shown that DDT can occur soon after the speed of a flame approaches the speed of sound.
2.3 Postulated episodic mechanism
The concept underlying episodic combustion originates from publications in the 1980s (Moore, 1983, 1987),
but did not progress due to the lack of experimental evidence. In the episodic form, it is postulated that thermal
radiation ignites particulate matter within a zone ahead of the flame. The simultaneous ignition of the particles
would then rapidly consume the fuel within a zone ahead of the flame. This then radiates further, repeating the
process in an episodic manner, with each rapid combustion generating pressures of several bar.
Attempts have been made to produce explosions by forward thermal radiation in ‘open’ environments but
without success. Dust and hybrid (dust with flammable gas) explosions do occur in different contexts (all
involving experiments in confined spaces) and numerous publications are available. However, none discuss
the episodic combustion concept.
In addition to the requirement for forward radiation from an explosion flame to ignite particulate matter in a
zone ahead of it, critically the explosion mechanism also needs to lift sufficient particulate matter into the
atmosphere ahead of the flame across large tracts of widely ranging ground types, including; concrete,
asphalt, gravel and soil.
Atkinson et al (2017a) show two images of flame propagation through clean pipes and those contaminated
with soot. It is implied that these support the episodic mechanism, but no evidence of ignition ahead of the
flame is demonstrated, only increased radiation by the soot after flame arrival.
The Buncefield JIP conducted independent large scale experiments to investigate the episodic mechanism
(Burgan 2014). They involved introduction carbon black into a flame venting from a confined explosion, giving
the explosion flame a high emissivity. This flame vented into a pre-mixed flammable 10x3x30m vapour cloud
with carbon black covering the floor and resting on elevated receptacles which dropped the carbon black into
the atmosphere ahead of the flame. The particulate matter was therefore introduced in a manner and a
quantity much more favourable to the proposed episodic mechanism than would occur in any real situation.
Despite having favourable conditions, an episodic explosion mechanism was not observed and no ignition
ahead of the flame front was occurred. The experiments are not referenced in Atkinson et al (2017a, 2017b).
In addition, investigations have been carried out at laboratory scale to define the conditions required to ignite
particulate matter ahead of a flame (Beyrau et al 2013, 2015, Li 2017). The experiments explored the
relationship between radiation intensity and material properties using a continuous wave laser. Possible
ignition was shown for a range of materials including different carbon blacks and non-reactive powders such
as iron and copper oxides. However, the required thermal radiation levels were well above the maximum
measured in the JIP large scale tests, even where carbon black was used to increase the flame emissivity.
In summary, there is no experimental evidence that any of the key elements of the mechanism would be
present in a real VCE. This lack of empirical evidence makes the proposition speculative at best. Suggesting
that it might be the cause of any VCE incident would only be justified if there were sound technical arguments
and, critically, if the existing understanding of explosions fails to explain the evidence. However, as already
shown, the well-defined and understood process of DDT is consistent with the evidence.
2.4 Recent alternative postulations
Thomas, 2018, proposes that the mechanism for generating pressures in the open car parks at Buncefield
involved interaction of shock waves with the combustion front. Whilst such interactions are known to intensify
combustion and play a part in the process of DDT, the suggested mechanism is not consistent with the
evidence for several reasons, including:
• The publication makes no reference to the Buncefield JIP experimental studies (Burgan 2014).
Consequently, it states that the best estimate of the overpressures in the open car parks and within
the cloud are 700-1000 mbar. As already discussed, this is incorrect.
• The mechanism would not explain the homogeneous level of damage throughout the vapour cloud.
• The mechanism would not reproduce the directional indicators observed at Buncefield.
It is surprising that the author was not aware of key experimental evidence developed by the JIP. Once this
evidence is added, the proposed mechanism fails the test of consistency with the evidence.
3. Evidence used to reject DDT
Atkinson et al 2017a, 2017b make a clear statement that there is evidence that some VCE incidents, including
Buncefield and Jaipur, did not involve a DDT or ‘standard’ deflagrations, justifying the episodic mechanism.
The ‘evidence’ that leads to the rejection of DDT mostly relates to the damage caused to items within a
detonating vapour cloud. For example, it is stated “Light weight steel elements in detonation tests showed
700
characteristic continuously curved shapes” when exposed to detonation pressures. It shows pictures of curved
steel angles that had formed theuprights of polythene support frameworks in detonation tests at DNV GL
Spadeadam Research and Testing. However, the picture shown was not typical of the damage; in most cases
there is no bending of the steel uprights (Johnson et al, 2018). Consequently, the lack of such deformation
cannot be used to exclude the possibility of a detonation.
Atkinson et al (2017a) show damage to oil drums from a detonation test, the Jaipur incident and a 2 bar static
test. The oil drum from Jaipur is shown on the right of Figure 1. The damage from the detonation test is taken
as representative of evidence of a detonation. Comparison is then made between the drum from Jaipur with
that from the static test, concluding that the Jaipur drum matched that from the static test, indicative of a
strong deflagration, not a detonation. However, a wider view of the area around the Jaipur drum shows
significant variability in damage patterns to other drums, most of which are not consistent with the static test.
Figure 3: Damage to oil drums immediately adjacent to that shown in the HSE 2017a (circled)
Atkinson et al (2017a) also fail to show examples of damage to two other Jaipur oil drums, one near full and
the other half full. Figure 4 shows these drums and a half full drum from a detonation test. The crushing of the
half full oil drum from the test is very similar to the half full drum at Jaipur. It is clear that the response of oil
drums varies significantly and that the conclusions drawn regarding exclusion of DDT cannot be sustained.
Figure 4: Recorded damage to oil drums at Jaipur (left and centre) and a drum from a detonation test (right)
Atkinson et al (2017a, 2017b) consider damage to other items including cars, buildings and instrument boxes.
Similar comment can be made to those above for all of these items, they do not allow dismissal of DDT.
4. Conclusions
The most important message that needs to be communicated is that there is clear and indisputable evidence
of high overpressures being generated throughout flammable vapour clouds in incidents, including in open
and uncongested areas.
701
Review of the evidence from Buncefield and Jaipur has shown:
• The standard interpretation of VCEs being only deflagration in congested process regions is not
consistent with the evidence from at least some VCE incidents.
• A scenario of a deflagration leading to DDT and the detonation continuing throughout the vapour
cloud has been tested against a wide range of evidence and found to be consistent.
• Evidence interpreted to exclude DDT in some VCEs (Atkinson et al 2017a, 2017b) including
Buncefield and Jaipur does not stand up to closer investigation.
There is no need to invoke a postulated new episodic explosion mechanism, particularly given the lack of
supporting evidence from experimental studies already carried out to investigate the mechanism. It is also
worth emphasising that a recent review of previous large VCEs has found no inconsistency in damage
evidence when deflagrative and detonative combustion is considered (Chamberlain et al, 2018).
References
Abdel-Gayed, R. G. and Bradley, D., 1982, The influence of turbulence upon the rate of turbulent burning.
Fuel-air explosions (ed. J. H. S. Lee & C. M. Guirao). Montreal: University of Waterloo Press
Atkinson, G., Hall, J., McGillivray, A., 2017a, Review of Vapour Cloud Explosion Incidents, Health and Safety
Executive, Research Report RR1113.
Atkinson, G., Cowpe, E., Halliday, J., Painter, D., 2017b, A review of very large vapour cloud explosions:
Cloud formation and severity, Journal of Loss Prevention in the Process Industries 48, 367-375.
Beyrau, F., Hadjipanayis, M. A., and Lindstedt, R. P., 2013, Ignition of fuel/air mixtures by radiatively heated
particles, Proceedings of the Combustion Institute, Vol: 34, Pages: 2065-2072, ISSN: 1540-7489.
Beyrau, F., Hadjipanayis, M. A., and Lindstedt, R. P., 2015, Time-resolved temperature measurements for
inert and reactive particles in explosive atmospheres, Proceedings of the Combustion Institute, Vol: 35,
Pages: 2067-2074, ISSN: 1540-7489.
Buncefield Major Incident Investigation Board, 2006, Initial Report to the Health and Safety Commission and
the Environment Agency of the investigation into the explosions and fires at the Buncefield oil storage and
transfer depot, Hemel Hempstead, on 11 December 2005.
Buncefield Major Incident Investigation Board, 2007, Explosion Mechanism Advisory Group report.
Burgan, B.A. 2014, Dispersion & Explosion Characteristics of Large Vapour Clouds, Steel Construction
Institute, Document ED024, 2015, available at http://www.fabig.com/video-publications/OtherPublications.
Chamberlain G., Oran E., Pekalski A. 2018, Detonations in industrial vapour cloud explosions, 12th
International Symposium on Hazards, Prevention and Mitigation of Industrial Explosions, Kansas City,
USA - August 12-17.
Davis S., Pagliaro J., Botwinick D., Engel D., van Wingerden K., Merilo E., Ziemba A., Groethe M., 2017,
Large Scale Testing Confirms: A Deflagration to Detonation Transition Needs to be Considered When
Facility Siting, 13th Global Congress on Process Safety, San Antonio.
Eyre, J.A. and Harrison, A.J 1987, The Effect of Obstacle Arrays on the Combustion of Large Pre-mixed
Gas/Air Clouds, Combustion Science and Technology, 52.
Harris R.J. and Wickens M.J. 1989, Understanding vapour cloud explosions – an experimental study, Institute
of Gas Engineers communication 1408.
Health and Safety Executive 2009, Buncefield Explosion Mechanism Phase 1, Research Report RR718.
Johnson, D.M., 2012, Characteristics of the Vapor Cloud Explosion Incident at the IOC Terminal in Jaipur,
29th October 2009, 8th Global Congress on Process Safety, Houston, TX.
Johnson D.M., Pekalski A., Tam, V.H.Y. and Burgan B.A., 2018, The importance of deflagration to detonation
transition in explaining major vapour cloud explosion incidents, ISHPMIE XII, Kansas City.
Kuhl, A.L., 1981, First Specialist Meeting (International), The Combustion Institute, Bordeaux, 1981.
Li, T. and Lindstedt, R.P. 2017, Thermal Radiation Induced Ignition of Multipoint Turbulent Explosions,
Process Safety and Environmental Protection. (4) 107, 108-122,
Moore, S.R. and Weinberg, F.J. 1983, A study of the role of radiative ignition in the propagation of large
explosions, Proc. R. Soc. Lond. A 385, 373–387
Moore, S.R. and Weinberg, F.J. 1987, Further studies of the role of radiative ignition in the propagation of
large explosions. Proc. R. Soc. Lond. A 409, 1–20.
Pekalski A., Puttock J., Chynoweth S. 2015, Deflagration to Detonation Transition in a Vapour Cloud
Explosion in Open but Congested Space: Large Scale Test, Journal of Loss Prevention in the Process
Industries, vol 36, p 365-370.
Thomas, G.O., 2018, Buncefield: A possible alternative multi-stage route to localised pressure generation,
Process Safety and Environmental Protection 119, 253-260.
702