You are on page 1of 59

Chapter 26

Case Histories

Chapter Outline
26.1 Introduction 448 26.5.7 The Worldwide Impact of the Bhopal
26.1.1 Incident Sources 448 Incident 470
26.1.2 Incident Databases 449 26.6 Pasadena 471
26.1.3 Reporting of Incidents 450 26.6.1 The Site and the Plant 471
26.1.4 Reporting of Injuries in Incidents 450 26.6.2 Events Prior to the Explosion 471
26.1.5 Reporting of Injuries at National Level 451 26.6.3 The Explosion 471
26.1.6 Incident Diagrams, Plans, and Maps 451 26.6.4 The Emergency and the Aftermath 472
26.1.7 Incidents Involving Fire Fighting 451 26.6.5 Some Lessons of Pasadena 472
26.1.8 Incidents Involving Condensed Phase 26.7 Canvey Reports 473
Explosives 451 26.7.1 First Canvey Report 474
26.1.9 Incidents Involving Spontaneously 26.7.2 First Canvey Report: Installations and
Combustible Substances 451 Activities 474
26.1.10 Case Histories 452 26.7.3 First Canvey Report: Identified Hazards 474
26.2 Flixborough 452 26.7.4 First Canvey Report: Failure and Event Data 476
26.2.1 The Site and the Works 453 26.7.5 First Canvey Report: Hazard Models and Risk
26.2.2 The Process and the Plant 453 Estimates 476
26.2.3 Events Prior to the Explosion 453 26.7.6 First Canvey Report: Assessed Risks and
26.2.4 The Explosion 454 Actions 477
26.2.5 Some Lessons of Flixborough 454 26.7.7 Second Canvey Report 479
26.3 Seveso 457 26.7.8 Second Canvey Report: Reassessed Risks and
26.3.1 The Site and the Works 457 Actions 480
26.3.2 The Process and the Plant 457 26.7.9 Second Canvey Report: Technical Aspects 480
26.3.3 TCDD and its Properties 458 26.8 Rijnmond Report 480
26.3.4 Events at Seveso 458 26.8.1 The Investigation 480
26.3.5 The Release 1 458 26.8.2 Installations and Activities 480
26.3.6 The Release 2 458 26.8.3 Event Data 480
26.3.7 The Later Aftermath, Contamination, and 26.8.4 Hazard Models 480
Decontamination 460 26.8.5 Population Characteristics 480
26.3.8 Some Lessons from Seveso 461 26.8.6 Mitigation of Exposure 482
26.4 Mexico City 462 26.8.7 Individual Assessments 482
26.4.1 The Site and the Plant 462 26.8.8 Assessed Risks 482
26.4.2 The Fire and Explosion 1 463 26.8.9 Remedial Measures 482
26.4.3 The Emergency 464 26.8.10 Critiques 482
26.4.4 The Fire and Explosion 2 2 464 26.9 San Carlos De La Rapita Disaster 482
26.4.5 Some Lessons of Mexico City 465 26.9.1 The Camp Site 483
26.5 Bhopal 465 26.9.2 The Road Tanker 483
26.5.1 The Site and the Works 465 26.9.3 The Fire and Explosion 483
26.5.2 The Process and the Plant 466 26.9.4 The Emergency and the Aftermath 483
26.5.3 MIC and its Properties 467 26.9.5 Lessons of San Carlos De La Rapita Disaster 483
26.5.4 Events Prior to the Release 467 26.10 Piper Alpha 483
26.5.5 The Release 467 26.10.1 The Company 484
26.5.6 Some Lessons Learned from the Bhopal 26.10.2 The Field and the Platform 484
Incident 468 26.10.3 Events Prior to the Explosion 484

Lees’ Process Safety Essentials.


© 2014 Elsevier Inc. All rights reserved. 447
448 Lees’ Process Safety Essentials

26.10.4 The Explosion, the Escalation, and the 26.14 BP America Refinery Explosion, Texas City, Texas,
Rescue 485 USA 493
26.10.5 Some Lessons of Piper Alpha 486 26.14.1 The Company 493
26.10.6 Recommendations on the Offshore Safety 26.14.2 Incident Description 493
Regime 486 26.14.3 Aftermath and Emergency 496
26.11 Three Mile Island 486 26.15 Buncefield 496
26.11.1 The Company 486 26.15.1 Description of the Incident 496
26.11.2 The Site and the Works 486 26.15.2 Causes of the Incident 496
26.11.3 The Process and the Plant 486 26.15.3 Lessons Learned from the Incident 497
26.11.4 Events Prior to the Excursion 487 26.16 Space Shuttle Columbia Disaster 497
26.11.5 The Excursion 1 487 26.16.1 Development of the Space Shuttle Program 497
26.11.6 The Emergency and the Aftermath 488 26.16.2 Columbia’s Final Flight 497
26.11.7 The Excursion 2 488 26.16.3 Accident Analysis 497
26.11.8 Some Lessons 488 26.16.4 Other Factors Considered 498
26.12 Chernobyl 488 26.16.5 The Accident’s Organizational Causes 498
26.12.1 The Site and the Works 489 26.16.6 Implications for the Future of Human Space
26.12.2 The Process and the Plant 489 Flight 498
26.12.3 Events Prior to the Release 489 26.16.7 Recommendations 498
26.12.4 The Release 490 26.17 Deepwater Horizon 499
26.12.5 The Emergency and the Immediate 26.17.1 The Companies 499
Aftermath 490 26.17.2 The Site and the Works 499
26.12.6 The Later Aftermath 491 26.17.3 Events Prior to the Explosions 499
26.12.7 Some Lessons of Chernobyl 491 26.17.4 Events Prior to April 19 499
26.13 Hurricanes Katrina and Rita 491 26.17.5 Events on April 19 and 20 500
26.13.1 Introduction 491 26.17.6 Lessons from the Deepwater Horizon
26.13.2 Hurricane Katrina 491 Incident 502
26.13.3 Hurricane Rita 492 26.17.7 Impact 502
26.13.4 Effect on the Industry 492 Acronyms 502
26.13.5 Lessons Learned 493 References 503
26.13.6 Recommendations 493

26.1. INTRODUCTION profile US chemical plant and refinery case histories. As


a result of these investigations, some Alerts focusing on
An essential feature of the learning process in safety and certain type of incidents have been developed. EPA reports
loss prevention is the study of case histories. are available at the following internet website http://www.epa.
gov/emergencies/publications.htm.
The US Chemical Safety and Hazard Investigation
26.1.1 Incident Sources Board (CSB) was created by an Act of the US Congress
The powerful impact of the internet has changed the way in the late 1990s and their mission is to promote the
we obtain our information on process safety, as well as prevention of chemical accidents. The focus of the CSB
everything else. In recent years, various governmental is on-site and off-site chemical safety, determining causes
agencies and private organizations have improved our and preventing chemical-related incidents, fatalities,
access to details of recent and past case histories. A num- injuries, property damage and enhancing environmental
ber of websites provide a narrative of the event, the protection. The CSB started investigations on selected
technical reasons for the accident, as well as the manage- important chemical plant and refinery incidents in about
ment system shortcomings and valuable lessons learned. 1998, and also provide reports on the internet at the fol-
Written reports sometimes supported by graphics and lowing web address: http://www.csb.gov/investigations/
photos are often published by various agencies and pro- default.aspx
vided on governmental websites. There are also university The Center for Chemical Process Safety (CCPS) of the
and corporate organizations that collect incidents and American Institute of Chemical Engineers started provid-
publish the information on their websites. ing brief one-page summaries of types of hazards in late
Since about 1996, the US Environmental Protection 2001. These high-impact messages were designed with
Agency (EPA), under the Chemical Emergency Preparedness the chemical process operator and others involved with the
and Prevention, has investigated and reported on many high manufacturing processes to focus quickly on an important
Chapter | 26 Case Histories 449

topic. This publication is entitled the Beacon and can be 1979; Slater, 1978; Lewis, 1980; Davenport, 1987; Lenoir
found at: www.aiche.org/ccps/safetybeacon.htm. and Davenport, 1993), LPG (L.N. Davis, 1979), instrumen-
The Mary Kay O’Connor Process Safety Center at tation (Doyle 1972a,b; Lees, 1976), transport (Haastrup and
Texas A&M University in College Station, Texas, is an Brockhoff, 1990), and pipelines (Riley, 1979). The series
international leader in process safety and is increasing in Safety Digest of Lessons Learned by the API gives case his-
stature each year. The Center is a rich source of process tories with accompanying analyses. Another such series is
safety information as well as documented case histories. published by the American Oil Company (Amoco) which
The Center’s web address is: http://process-safety.tamu. includes Hazards of Water and Hazards of Air. The Canvey
edu. Reports and the Rijnmond Report gives information on cer-
The British Institution of Chemical Engineers tain incidents. The reports of the HSE provide case histories
(IChemE) offers links to a comprehensive list of well of investigations by a regulator, as do those of the NTSB,
over 150 other websites of incident reports. The IChemE which deal with rail, road, pipeline, and marine accidents.
site mainly describes incidents of interest to chemical or Incidents involving large economic loss are described in the
process engineers, and includes a time span of incidents periodic review 100 Large Losses by Marsh and
from historical to today’s events. The collection includes McLennan.
both information and viewpoint. The address of this site Case histories are also given in the Annual Report of
is: http://slp.icheme.org/incidents.html. HM Chief Inspector of Factories, The Chemical Safety
The Delft University of Technology of the Netherlands Summary of the Chemical Industry Safety and Health
offers a vivid array of about two dozen outstanding fire Council (CISHC), and in journals such as Petroleum
and explosion photos, an accident database, and more. The Review and NFPA Journal. In addition, the Loss Prevention
university’s main research effort is focused on measuring Bulletin issued by the IChemE is a major source of safety
and modeling dust and gas explosions. The incident photos case studies for the process industries. There are also
can be found at: http://www.dct.tudelft.nl/part/explosion/ numerous case histories described in much greater detail in
gallery.html. various papers and reports. The analysis of case histories
Another internet source of process safety information, to draw the relevant lessons is exemplified pre-eminently
including case histories, is www.acusafe.com. The AcuSafe in the work of Trevor Kletz’s many books and numerous
website provides a comprehensive electronic newsletter technical articles:
entitled AcuSafe News. Their newsletter is devoted to serv- G Critical Aspects of Safety and Loss Prevention (1990),
ing industry, government, and interested members of the G Lessons from Disaster—How Organisations Have No
public by providing process safety practices, incident news,
Memory and Accidents Recur (1993)
lessons learned, and regulatory developments. G What Went Wrong?—Case Histories of Process Plant
A booklet entitled Large Property Damage Losses
Disasters, 5th edn (2009)
in the Hydrocarbon-Chemical Industry—A Thirty Year G An Engineer’s View of Human Error, third edition
Review is a source of information in print form. This
(2001a)
publication presents summaries of large property losses G Learning from Accidents, third edition (2001b) and
of incidents from around the globe. The information is
numerous technical articles.
from a property insurance viewpoint and is presented in G Still Going Wrong!: Case Histories of Process Plant
an easy-to-understand form. Most of the damages are the
Disasters and How They Could Have Been Avoided
result of fires and explosions occurring in refineries, pet-
(2003)
rochemical plants, gas-processing plants, terminals, and
offshore, with a miscellaneous category also provided.
It is updated every few years. The 21st edition of Marsh’s 26.1.2 Incident Databases
The 100 Largest Property Losses 1972 2009 (in the
There are a number of databases specifically dealing with
Hydrocarbon Industry) reviews the 100 largest property
case histories. They include the following:
damage losses that have occurred in the hydrocarbon
processing industries since 1972. G The incident databases MHIDAS (the Major Hazard
Collections of case histories have been published in Incident Data Service) is a database of incidents involv-
Case Histories of Accidents in the Chemical Industry by the ing hazardous materials that had an off-site impact, or
Manufacturing Chemists Association (1962 , c1975.) had the potential to have an off-site impact. MHIDAS
and by Kier and Muller (1983). Specialized collections is maintained by AEA Technology plc, on behalf of the
have been published on particular topics such as major UK Health and Safety Executive and EIDAS (Explosives
hazards (Harvey, 1979, 1984; Carson and Mumford, 1979; Incidents Database Advisory Service), which contains
Lees, 1980; Marshall, 1987), fire and explosions (Vervalin, readily accessible and searchable information about the
1964, 1973; Doyle, 1969), vapor cloud explosions (Gugan, causes and effects of explosives and firework incidents
450 Lees’ Process Safety Essentials

in the United Kingdom, from the 1850s up to the present accident of the magnitude of Bhopal is not reported in
day, maintained by HSE. Website: http://www.hse.gov. countries with a free press is negligible. However, as the
uk/explosives/eidas.htm. size of the incident decreases, the probability that it is not
G TNO has developed the FACTS incident database reported or at least is not picked up in incident collections
covering 24,000 (industrial) accidents (incidents) and databases increases.
involving hazardous materials or dangerous goods that OSHA (Regulations) Part 1094: recording and reporting
have happened worldwide over the past 90 years. occupational injuries and illness was developed to require
G Major Accident Reporting System (MARS) is a distri- employers to report and record work-related fatalities, inju-
bution information network, consisting of 15 local ries, and illness. The standard also provides guidelines about
databases from the European Union and the European the process that should be performed. For more information,
Commission’s Joint Research Centre in Ispra (MAHB). refer to OSHA’s website: http://www.osha.gov/pls/oshaweb/
The database allows complex text retrieval and pattern owasrch.search_form?p_doc_type5STANDARDS&p_toc_
analysis. level51&p_keyvalue51904.
G Hydrocarbon Releases Database System contains infor- Key highlights of Standard part 1904:
mation, dating from October 1992, concerning offshore G All employers covered by the OSH Act must report to
releases of hydrocarbons reported to the HSE Offshore
OSHA any workplace incident that results in a fatality
Division (OSD) under the Reporting of Injuries,
or the hospitalization of three or more employees
Diseases and Dangerous Occurrences Regulations 1995
[1904.39].
(RIDDOR), and prior offshore legislation. Website: G Each employer is required by Part 1904.4 to keep records
https://www.hse.gov.uk/hcr3/.
G
of fatalities, injuries, and illnesses, and must record each
HSEES: US federal database. The activities of the HSEES
fatality, injury, and illness that is work-related, is a new
database program are now administered by the National
case, and meets one or more of the general recording
Toxic Substance Incidents Program (NTSIP) Website:
criteria of part 1904.7, or if it applies to specific cases in
http://www.atsdr.cdc.gov/ntsip/National_Database.html.
G
part 1904.8 through part 1904.12.
ARIA (France) database records accidents which G Employers must save the OSHA 300 Log, the privacy
had or could have had an adverse effect on public
case list (if one exists), the annual summary, and the
safety or health, agriculture, nature, and the environ-
OSHA 301 Incident Report forms for 5 years following
ment. Website: http://www.aria.developpement-durable.
the end of the calendar year that these records cover
gouv.fr/The-ARIA-Database--5425.html.
G
[Part 1904.33].
Fire & Explosion in the Canadian Upstream Oil and
Gas Industry. Website: http://www.firesandexplosions. Within 8 h after the death of any employee from a
ca/index.php. work-related incident or the in-patient hospitalization
of three or more employees as a result of a work-related
incident, employers must orally report the fatality/
26.1.3 Reporting of Incidents multiple hospitalization by telephone or in person to the
The extent and accuracy of the reporting of incidents and Area Office for OSHA, and the US Department of Labor
injuries is variable and this creates problems, particularly that is nearest to the site of the incident [part 1904.39]
for attempts to perform statistical analysis of incident (OSHA).
data.
Three distinct problems may be identified: (1) occur-
rence of an incident, (2) injuries associated with an incident, 26.1.4 Reporting of Injuries in Incidents
and (3) national injury statistics. These three cases are
considered in this section and in the next two sections, For various reasons, accounts of incidents tend to differ
respectively. in the number of injuries and, to a lesser extent, fatalities
The awareness of incidents in the engineering commu- that are reported. A discussion of this problem has been
nity worldwide varies according to the country in which given by Haastrup and Brockhoff (1991).
the incident has occurred and the size and impact of the There are a number of reasons for differences in the
incident. For example, in the recent past, incidents in the numbers quoted. One is that early reports of an incident
United States have generally been reported in the techni- tend not be very accurate, but are sometimes quoted with-
cal press, but reports of comparable incidents in the out sufficient qualification and the numbers then receive
USSR have been relatively few. currency.
With regard to the effects of scale and impact, the With regard to fatalities, a difference arises between
probability of worldwide reporting of an incident clearly immediate and delayed deaths. A case where a proportion
increases with these factors. The probability that an of delayed deaths is fairly common is burn casualties.
Chapter | 26 Case Histories 451

The most frequent and large differences, however, are missiles, and location of victims. In particular, such
in ‘injuries’. Here, much of the difference can usually be diagrams are a feature of the accident reports by the HSE
accounted for by differences in definition. and the NTSB and of case histories described in the Loss
As an illustration, consider the injuries in the explosion Prevention Bulletin.
at Laurel, Mississippi, on January 25, 1969. The NTSB
report on this incident (NTSB 1969 RAR) states that 2 per-
sons died, 33 received treatment in hospital, and numerous
26.1.7 Incidents Involving Fire Fighting
others were given first aid. Some authors have therefore A feature of some interest in incidents is the experience
quoted this as 2 dead, 33 injured. Eisenberg et al. (1975) gained in fire fighting. Further information may be found
refer to the NTSB report and also to a private communica- in the following accounts:
tion from a railroad source and state that 976 persons were
Brindisi, 1977 (Mahoney, 1990)
injured, 17 being in hospital for more than a month. This
Milford Haven, 1983 (Dyfed County Fire Brigade, 1983)
incident is one of those quoted by Haastrup and Brockhoff
Thessalonika, 1986 (Browning and Searson, 1989)
(1991) as an example of the problem.
Grangemouth, 1987 (HSE, 1989)
Port Heriot, 1987 (Mansot, 1989)
26.1.5 Reporting of Injuries
at National Level
26.1.8 Incidents Involving Condensed
It is normal for there to be a regulatory requirement for the Phase Explosives
reporting, as a minimum, of deaths and injuries. In the
United Kingdom, the relevant regulations are RIDDOR A number of the incidents given in the following sections
1985. The information gathered in this way is published by involve explosives. There have been, however, a large
the HSE in the series Health and Safety Statistics. number of other explosives and munitions incidents that
The reporting is incomplete. A study supplementary to are of only marginal interest here. An account of explo-
the household-based Labour Force Survey 1990 in the sions up to 1930 is given in History of Explosions by
United Kingdom showed that only approximately 30% of Assheton (1930), later treatments are in Explosions in
non-fatal injuries were reported to the HSE, but that the History by Wilkinson (1966), and Darkest Hours by Nash
level of reporting varied significantly across industries (1976). Some principal explosions are listed by Nash. His
(Kiernan, 1992). For the energy industry, the proportion list gives death tolls that in some cases differ from those
of such accidents reported was 75%, and it seems proba- given elsewhere.
ble that the level of reporting in process industries such
as oil refining, petrochemical, and chemicals is similar. 26.1.9 Incidents Involving Spontaneously
In the United States, the Hazardous Substances
Combustible Substances
Emergency Events Surveillance (HSEES) system, maintained
by the Agency for Toxic Substances and Disease Registry Spontaneous combustible substances are materials which
(ATSDR), is a key component in monitoring the acute health can ignite without any flame, spark, heat, or other ignition
effects, causes, and circumstances of chemical, biological, source. Hence, the definition of ‘spontaneous combustion’
radiological, and nuclear release in the United States is ‘combustion that results when materials undergo atmo-
(MKOPSC, 2009). For all incidents, other than petroleum- spheric oxidation at such a rate that the heat generation
only releases, the HSEES system details the substances exceeds heat dissipation and the heat gradually builds up
involved, causes of the incident, associated equipment to a sufficient degree to cause the mass of material to
items, the type of location, victim demographics, the type of inflame’ (Carson and Mumford, 2002). A number of
emergency response, injury details, personnel protective materials can be classified as spontaneous combustibles
equipment in use, nearby vulnerable populations, and other including linseed oil, alkyd enamel resins, and drying
pertinent information. Currently, 14 state health departments oils. A more comprehensive list of spontaneous combusti-
have cooperative agreements with ATSDR to participate bles can be found in the ‘National Fire Protection
in HSEES: Colorado, Florida, Iowa, Louisiana, Michigan, Handbook (18th edition), Table A.10: Materials subject
Minnesota, New Jersey, New York, North Carolina, Oregon, to spontaneous heating’. The worst consequences of
Texas, Utah, Washington, and Wisconsin. spontaneous combustion include fatalities through blast
trauma and asphyxiation from CO and mental disorders in
survivors of disaster/accidents.
26.1.6 Incident Diagrams, Plans, and Maps A large number of incidents involving spontaneous
Many accounts of incidents give diagrams, plans, or maps combustion happened in the underground mining environ-
showing features such as derailed tank cars, location of ment. For example, there has been an increase in the number
452 Lees’ Process Safety Essentials

of spontaneous combustion incidents in the past years,


culminating in the closure of Southland Colliery in TABLE 26.1 The Most Frequently Mentioned
December 2003. Since 1972, spontaneous combustion Chemicals
has resulted in three underground mine explosions which Ammonia 48 Chlorine 37
killed 41 workers in Queensland and some pit closures in
Caustic soda 88 Sulfuric acid 46
New South Wales (Ham, 2005). A report by Richardson
and Ham (1996) outlines several incidents of spontane-
ous combustion in Queensland underground coal mines.
A total of 51 spontaneous combustion incidents were
reported in Queensland between 1972 and 2004. The
most damaging of these incidents were Box Flat in 1972, TABLE 26.2 The Most Frequently Mentioned
Kianga in 1975, and Moura No 2 Mine in 1994. These Operations
incidents involved explosions that resulted in 17, 13,
Loading 38 Steaming 32
and 11 fatalities, respectively, as well as closures of the
mines. Maintenance 82 Tank entry 39

Pipe fitting 27 Transfer 46


Process reaction 66 Unloading 64
26.1.10 Case Histories Sampling 24 Welding 36
One of the principal sources of case histories is the MCA
collection. There are a number of themes which recur
repeatedly in these case histories. They include:
Failure of communications
Failure to provide adequate procedures and instructions TABLE 26.3 The Most Frequently Mentioned
Failure to follow specified procedures and instructions Equipment
Failure to follow permit-to-work systems
Centrifuge 24 Machine 27
Failure to wear adequate protective clothing
Failure to identify correctly plant on which work is to Cylinder 26 Pump 60
be done Drum 76 Rotameter, sight glass 27
Failure to isolate plant, to isolate machinery and
Hose 57 Tank 46
secure equipment
Failure to release pressure from plant on which work Industrial truck 31 Tank car 59
is to be done Laboratory 85 a
Tank truck 35
Failure to remove flammable or toxic materials from
Line 105 Valve 86
plant on which work is to be done
a
Failure of instrumentation In the first 1623 case histories.
Failure of rotameters and sight glasses
Failure of hoses
Failure of, and problems with, valves
Incidents involving exothermic mixing and reaction
processes 26.2. FLIXBOROUGH
Incidents involving static electricity
At about 4:53 p.m. on Saturday, June 1, 1974, the
Incidents involving inert gas.
Flixborough Works of Nypro (UK) Ltd (Nypro) were
In the 2108 case histories described, the most fre- virtually demolished by an explosion of war-like dimen-
quently mentioned chemicals and the corresponding num- sions. Of those working on the site at the time, 28 were
ber of entries are given in Table 26.1. killed and 36 others suffered injuries. If the explosion had
The most frequently mentioned operations are given occurred on an ordinary working day, the number of
in Table 26.2. casualties would have been much greater. Outside the
The most frequently mentioned kinds of equipment Works, injuries and damage were widespread, but no one
are given in Table 26.3. was killed. Fifty-three people were recorded as casualties
It is emphasized, however, that in many instances it is by the casualty bureau, which was set up by the police;
not appropriate to assign a single cause. For more details, hundreds more suffered relatively minor injuries, which
refer to MCA collections. were not recorded. Property damage extended over a
Chapter | 26 Case Histories 453

wide area, and a preliminary survey showed that 1821 reactor were a pressure of 8.8 kg/cm2 and a temperature
houses and 167 shops and factories had suffered to a of 155 C. The reaction is exothermic.
greater or lesser degree (Parker, 1975—the Flixborough The heat required for initial warm-up and for supple-
Report, para. 1). mentation of the heat of reaction during normal operation
The Flixborough explosion was by far the most seri- was provided by a steam-heated heat exchanger on the
ous accident that had occurred in the chemical industry in reactor feed. The steam flow to the exchanger was con-
the United Kingdom for many years. trolled by an automatic control valve. There was a bypass
Within a month of the disaster, a Court of Inquiry around this valve, which was needed to pass the larger
under the chairmanship of Mr. Parker was set up under quantities of steam required during start-up.
Section 84 of the Factories Act 1961 to establish the Removal of the heat of reaction from the reactors dur-
causes and circumstances of the disaster and to point out ing normal operation was effected by vaporizing part of
any lessons which might be learned. the cyclohexane liquid. The vaporized cyclohexane passed
The Court’s report The Flixborough Disaster, Report out in the off-gas from the reactors. The rest of this off-gas
of the Court of Inquiry (Parker, 1975) (the Flixborough was mainly nitrogen with some unreacted oxygen.
Report), was the most comprehensive inquiry conducted The off-gas passed through the feed heat exchanger
in the United Kingdom at that time of a disaster in the and then through a cooling scrubber and an absorber, in
chemical industry. which the cyclohexane was condensed out, and thence via
The Flixborough disaster was of crucial importance in an automatic control valve to a flare stack.
the development of safety and loss prevention in the The atmosphere in the reactor was controlled using
United Kingdom. It made both the industry and the public nitrogen from high-pressure nitrogen storage tanks. The
much more aware of the potential hazard of large chemi- nitrogen was brought into the works by tankers.
cal plants and led to an intensification of both the efforts The reactor pressure was controlled by manipulating
within industry to ensure the safety of major hazard plants the control valve on the off-gas line. Safety valves venting
and of the demands for public controls on such plants. into the relief header to the flare stack were set to open
The setting up of the Advisory Committee on Major at 11 kgf/cm2.
Hazards (ACMH) at the end of 1974 was a direct result A trip system was provided which shut off air to, and
of the Flixborough disaster. injected nitrogen into, the reactors in the event of either a
high oxygen content in the off-gas or a low liquid level in
the nitrogen supply tank. This trip could be disarmed;
26.2.1 The Site and the Works however, by setting the timer to zero, fixing the duration
of the purge.
The nearest villages are Flixborough itself and Amcotts,
both of which are about half a mile away. The town of
Scunthorpe lies at a distance of approximately 3 miles. 26.2.3 Events Prior to the Explosion
The works is surrounded by fields, and the population
On the evening of March 27, 1974, it was discovered
density in the neighborhood beyond is very low.
that Reactor No. 5 was leaking cyclohexane. The reactor
Other plants on the site included an acid plant and a
was constructed of 1/2 in. mild steel plate with 1/8 in.
hydrogen plant. There was also a large ammonia storage
stainless steel bonded to it on the inside. A vertical crack
sphere.
was found in the mild steel outer layer of the reactor. The
leakage of cyclohexane from the crack indicated that
the inner stainless steel layer was also defective. It was
26.2.2 The Process and the Plant decided that the plant be shut down for a full investigation.
The cyclohexane plant consisted of a train of six reactors The following morning’s inspection revealed that the crack
in series in which cyclohexane was oxidized to cyclohex- had extended some 6 ft. This was a serious state of affairs
anone and cyclohexanol by air injection in the presence and a meeting was called to decide on a plan of action.
of a catalyst. The feed to the reactors was a mixture of The decision was made to remove Reactor No. 5 and to
fresh cyclohexane and recycled material. The product install a bypass assembly to connect Reactors No. 4 and 6
from the reactors still contained approximately 94% of so that the plant could continue production.
cyclohexane. The liquid reactants flowed from one reac- The openings to be connected on these reactors were
tor to the next by gravity. In subsequent stages, the reac- 28 in. diameter, with bellows on the nozzle stubs, but the
tion product was distilled to separate the unreacted largest pipe which was available on site and which might
cyclohexane, which was recycled to the reactors, and the be suitable for the bypass was 20 in. diameter. The two
cyclohexanone and cyclohexanol, which were converted flanges were at different heights so that the connection
to caprolactam. The design operating conditions in the had to take the form of a dog-leg of three lengths of
454 Lees’ Process Safety Essentials

20 in. pipe welded together with flanges at each end with or without contribution from a fire on a nearby 8 in.
bolted to the existing flanges on the stub pipes on the pipe. The cyclohexane formed a vapor cloud and the
reactors. flammable mixture found a source of ignition.
Calculations were done to check (1) that the pipe was At about 4:53 p.m., there was a massive vapor cloud
large enough for the required flow and (2) that it was explosion. The explosion caused extensive damage and
capable of withstanding the pressure as a straight pipe. started numerous fires. The blast of the explosion shat-
No calculations were made which took into account the tered the windows of the control room and caused the
forces arising from the dog-leg shape of the pipe. control room roof to collapse. Of the 28 people who died
No drawing of the bypass pipe was made other than in in the explosion, 18 were in the control room. Some of
chalk on the workshop floor. the bodies had suffered severe injuries from flying glass.
The bypass assembly was supported by a scaffolding Others were crushed by the roof. No one escaped from
structure. This scaffolding was intended to support the the control room.
pipe and to avoid straining of the bellows during con- The main office block was also demolished by the
struction of the bypass. It was not suitable as a permanent blast of the explosion. Since the accident occurred on a
support for the bypass assembly during normal operation. Saturday afternoon, the offices were not occupied. If they
No pressure testing was carried out either on the pipe had been, the death toll would have been much higher.
or on the complete assembly before it was fitted. A pressure The fires on the site burned for many days. Even after
test was performed on the plant, however, after installation 10 days the fires were hindering rescue work on the site.
of the bypass. The equipment was tested to a pressure The large ammonia sphere was lifted up a few inches.
of 9 kgf/cm2, but not up to the safety valve pressure of It leaked slightly at a flange, but the leak was not serious.
11 kgf/cm2. The test was pneumatic, not hydraulic.
Following these modifications, the plant was started
up again. The bypass assembly gave no trouble. There 26.2.5 Some Lessons of Flixborough
did appear, however, to be an unusually large usage of There are numerous lessons to be learned from the
nitrogen on the plant, and this was being investigated at Flixborough disaster. The lessons include both public
the time of the accident. controls on major hazard installations and the manage-
On May 29, the bottom isolating valve on a sight glass ment of such installations by industry. In the latter area,
on one of the vessels was found to be leaking. It was there are lessons on both management systems and tech-
decided to shut the plant down to repair the leak. nological matters and on both design and operational
On the morning of June 1, start-up began. The precise aspects.
sequence of events is complex and uncertain. The crucial The fact that alternative hypotheses were advanced
feature, however, is that the reactors were subjected to concerning the cause of the explosion does not detract
a pressure somewhat greater than the normal operating from these lessons, but rather means that a greater num-
pressure of 8.8 kgf/cm2. ber of lessons can be drawn. Some of these lessons are
A sudden rise in pressure up to 8.5 kgf/cm2 occurred now considered.
early in the morning when the temperature in Reactor No. 1
was still only 110 C and that in the other reactors was
26.2.5.1 Public Controls on Major Hazard
less, while later in the morning, when the temperature in the
Installations
reactors was closer to the normal operating value, the
pressure reached 9.1 9.2 kgf/cm2. The effect of the Flixborough disaster was to raise the
The control of pressure in the reactors could normally general level of awareness of the hazard from chemical
be affected by venting the off-gas, but this procedure plants and to make the existing arrangements for the con-
involved the loss of considerable quantities of nitrogen. trol of major hazard installations appears inadequate.
Shortly after warm-up began, it was found that there was The government therefore set up the ACMH to advise
insufficient nitrogen to begin oxidation and that further means of control for such installations. The committee
supplies would not arrive until after midnight. Under issued three reports (Harvey, 1976, 1979, 1984).
these circumstances, the need to conserve nitrogen would This work was a major input to the development of
tend to inhibit reduction of pressure by venting. the EC Major Accident Hazards Directive, which was
implemented in the United Kingdom as the CIMAH
Regulations 1984. These require the operator of a major
26.2.4 The Explosion hazard installation to produce a safety case. Major hazard
During the late afternoon, an event occurred which installations receive a greater degree of supervision by
resulted in the escape of a large quantity of cyclohexane. the local Factory Inspectorate. The CIMAH safety case
This event was the rupture of the 20 in. bypass system, plays an important part in this.
Chapter | 26 Case Histories 455

26.2.5.2 Siting of Major Hazard Installations The use of a comprehensive set of procedures is
another important aspect of the management system.
As the Flixborough Report (para. 11) points out, the
A crucial procedure which was deficient at Flixborough
number of casualties from the explosion might have been
was that for the control of plant modifications.
much greater if the site had not been on open land.
The role of the safety officer at Flixborough was not
The question of the siting of major hazard installations,
well defined.
or more generally land use planning in relation to such
The importance for major hazard installations of the
installations, became a principal concern of the ACMH,
management and the management system was the single
and is treated in its three reports.
most prominent theme in the work of the ACMH.
Emphasis by the HSE on management aspects has
26.2.5.3 Licensing of Storage of steadily grown.
Hazardous Materials
The situation at Flixborough revealed the need for better 26.2.5.6 Relative Priority of Safety
methods of notification of major hazard installations to and Production
the local planning authorities and for greater guidance
to these authorities by the HSE. The notification of major The Flixborough Report (paras 57, 206) drew attention to
hazard installations is a requirement of the 1982 NIHHS the conflict of priorities between safety and production.
Regulations and the 1984 CIMAH Regulations. It states:
We entirely absolve all persons from any suggestion that their
26.2.5.4 Regulations for Pressure Vessels desire to resume production caused them knowingly to embark
and Systems on a hazardous course in disregard of the safety of those operat-
ing the Works. We have no doubt, however, that it was this
The escape of cyclohexane at Flixborough was caused by desire which led them to overlook the fact that it was potentially
a failure of the integrity of a pressure system. hazardous to resume production without examining the remain-
At the time of the Flixborough incident, legislation on ing reactors and ascertaining the cause of the failure of the fifth
pressure systems in the United Kingdom consisted of reg- reactor. We have equally no doubt that the failure to appreciate
ulations for steam boilers and receivers and air receivers. that the connection of Reactor No. 4 to Reactor No. 6 involved
It failed to cover the Flixborough situation in two crucial engineering problems was largely due to the same desire.
respects. It applied only to steam and air, not to hazardous
materials such as cyclohexane, and it dealt only with
pressure vessels, not with pressure systems. This latter
point is relevant, because at Flixborough, the failure 26.2.5.7 Use of Standards and Codes
occurred in a pipe not a vessel. of Practice
The Flixborough Report (para. 209) recommends that As the Flixborough Report (paras 61 73) describes, the
existing regulations relating to the modification of steam 20 in. bypass assembly was not constructed and installed
boilers should be extended to apply to pressure systems in accordance with the relevant standards and codes of
containing hazardous materials. practice. The bellows manufacturer, Teddington Bellows
Ltd, produced a Designer’s Guide which made it clear
26.2.5.5 The Management System for Major that two bellows should not be used out of line in the
Hazard Installations same pipe without adequate support for the pipe.

The works did not have a sufficient complement of


qualified and experienced people. Consequently, manage-
26.2.5.8 Limitation of Inventory in the Plant
ment was not able to observe the requirement that persons
given responsibilities should be competent to carry them The Flixborough Report (para. 14) makes it clear that the
out. In particular, there was no works engineer in post large inventory of flammable material in the plant contrib-
and no adequately qualified mechanical engineer on site. uted to the scale of the disaster.
Moreover, individuals tended to be overworked and thus The Second Report of the ACMH proposes that limita-
more liable to make errors. tion of inventory should be taken as a specific design
The management system, however, is more than the objective in major hazard installations.
individuals. It includes the whole structure which supports The limitation of inventory is a particular aspect of
them. Thus the system should provide, for example, for the the more general principle of inherently safer design,
coverage of absence due to resignation, illness, and so on. which is now widely recognized.
456 Lees’ Process Safety Essentials

26.2.5.9 Engineering of Plants 26.2.5.14 Decision Making Under


for High Reliability Operational Stress
The explosion at Flixborough occurred during a plant The Flixborough Report (para. 205) draws attention to
start-up. The Flixborough Report (para. 206) suggests the problem of decision making under operational stress
that special attention should be given to factors which and emphasizes the desirability of reducing the number of
necessitate the shut-down of the chemical plant so as to critical management decisions which have to be made
minimize the number of shut-down/start-up sequences under these conditions.
and to reduce the frequency of critical management Such critical decisions are not necessarily confined
decisions. to management, however. Process operators may also be
required to take important decisions in emergency
conditions.
26.2.5.10 Dependability of Utilities
The high-pressure nitrogen required for the blanketing
of the reactors at Flixborough was brought into the works 26.2.5.15 Restart of Plant After Discovery
by tankers. There was insufficient nitrogen available of a Defect
during the start-up when the explosion occurred. The Following the discovery of the serious defect in Reactor
Flixborough Report (para. 211) emphasizes the impor- No. 5 at Flixborough, the reactor was removed, a bypass
tance of assuring dependable supplies of nitrogen where assembly was installed, and the plant was started up
these are necessary for safety. again. The Flixborough Report (para. 57) is critical of the
fact that the remaining reactors were not examined, and
the cause of the failure in the fifth reactor was not ascer-
26.2.5.11 Limitation of Exposure of Personnel
tained before plant start-up.
The Flixborough Report (para. 1) states that the number
of casualties would have been much greater if the explo-
sion had occurred on a weekday instead of on a Saturday. 26.2.5.16 Control of Plant and Process
The First Report of the ACMH (para. 68) suggests that Modifications
limitation of exposure of personnel be made a specific
The Flixborough Report (para. 209) states:
design objective.
Aspects of limitation of exposure are controls on access The disaster was caused by the introduction into a well-designed
to hazardous areas and design and location of buildings in and constructed plant of a modification which destroyed its integ-
or near such areas. rity. The immediate lesson to be learned is that measures must be
taken to ensure that the technical integrity of plant is not violated.

26.2.5.12 Design and Location of Control As it happens, there was also a process modification
Rooms and Other Buildings at Flixborough, although this is not emphasized in the
report. The agitator in Reactor No. 4 was not in use at
Of the 28 deaths at Flixborough 18 occurred in the con- the time of the disaster. The absence of agitation in the
trol room. The Flixborough Report (para. 218) refers to reactor could have allowed a water layer to accumulate
various suggestions made to the inquiry concerning the more easily in the bottom of the vessel.
siting of control rooms, laboratories, offices, etc., and the
construction of control rooms on blockhouse principles.
The construction of buildings for chemical plant is 26.2.5.17 Security of and Control of Access
considered in the First Report of the ACMH (para. 69). to Plant
The Flixborough Report (para. 194) draws attention to the
26.2.5.13 Control and Instrumentation of Plant fact that there were two unguarded gates through which
it was possible for anyone at any time to gain access,
The control and instrumentation system was not a promi-
although this fact did not contribute to the disaster.
nent feature in the Flixborough inquiry. The Flixborough
Report (para. 204) considered that the controls in the
control room followed normal practice. But it also states
26.2.5.18 Planning for Emergencies
‘Nevertheless we conclude from the evidence that greater
attention to the ergonomics of plant design could provide The Flixborough Report (para. 222) calls for a disaster
rewarding results’. plan for major hazard installations.
Chapter | 26 Case Histories 457

26.2.5.19 The Metallurgical Phenomena Milan. A white cloud drifted from the works and material
from it settled out downwind. Among the substances
The Flixborough disaster drew attention to several important
deposited was a very small amount of TCDD, one of the
metallurgical phenomena. Thus, the Flixborough Report
most toxic chemicals known. There followed a period of
describes the nitrate stress corrosion cracking of mild steel
great confusion due to lack of communication between
(paras 53, 212); the creep cavitation of stainless steel (para.
the company and the authorities and the latter’s inexperi-
214); the zinc embrittlement of stainless steel (para. 213);
ence in dealing with this kind of situation. Over the next
and the use of clad mild steel vessels (para. 224).
few days in the contaminated area, animals died and
The HSE subsequently issued Technical Data Notes
people fell ill. A partial and delayed evacuation was car-
(1976 TON 53/1, 53/2; 1977 TON 53/3) on the first three
ried out. In the immediate aftermath, there were no deaths
of these problems.
directly attributable to TCDD, but a number of pregnant
It was apparent from the discussion in the engineering
women who had been exposed had abortions.
profession following publication of the report that although
A Parliamentary Commission of Inquiry, drawn
metallurgical specialists were aware of failure due to zinc
equally from the Chamber of Deputies and the Senate
embrittlement, the phenomenon was not well known
and chaired by Deputy B. Orsini, was set up. The
among engineers generally.
Commission’s report (the Seveso Report) (Orsini, 1977,
1980) is a far-ranging inquiry not only into the disaster
26.2.5.20 Vapor Cloud Explosions but also into controls over the chemical industry in Italy.
The explosion at Flixborough was a large vapor cloud The impact of the Seveso disaster in Continental
explosion. Although such explosions had become more Europe has in some ways exceeded that of Flixborough
common in the preceding years, none compared with and has led to much greater awareness of process industry
Flixborough in scale and impact. The Flixborough Report hazards on the part of the public and demands for more
(para. 215) draws attention to the marked lack of informa- effective controls.
tion on the conditions under which a vapor cloud can The EC Directive on Major Accident Hazards of 1982
explode. was a direct result of the Seveso disaster, and indeed this
Directive was initially often referred to as the ‘Seveso
26.2.5.21 Investigation of Disasters Directive’.
In addition to the Seveso Report, there are other
and Feedback of Information on
accounts of the Seveso disaster and of TCDD in The
Technical Incidents Superpoison by Margerison et al. (1980) and The Chemical
The Flixborough disaster was investigated by a Court of Scythe by Hay (1976a e, 1981, 1982), Bolton (1978),
Inquiry. There was some feeling in the engineering pro- Marshall (1980), Theofanous (1981, 1983), Rice (1982),
fession that a legal inquiry of this kind is not a satisfac- Sambeth (1983), and Howard (1985).
tory means of establishing the facts concerning technical
incidents.
The Flixborough Report (para. 216) states that the 26.3.1 The Site and the Works
inquiry would have been greatly assisted if the essential
When ICMESA built its works at Meda, the site was sur-
instrument records in the control room had not been
rounded by fields and woods. Over the years, however,
destroyed in the explosion, and recommends that consid-
the area near the site was developed.
eration be given to systems that record and preserve
The reactor, Reactor A101, in which the runaway
vital plant information, such as the ‘black box’ used in
occurred, was in Department B of the works.
aircrafts.
The points just outlined by no means exhaust the
lessons to be learned from the Flixborough disaster. In
26.3.2 The Process and the Plant
particular, there are many instructive aspects of the 8 in.
pipe hypothesis relating to such features as lagging fires, The process which gave rise to the accident was the
directed flames, and sprinkler sensor performance. production of 2,4,5-trichlorophenol (TCP) in a batch
reactor.
TCP is used for herbicides and antiseptics. Givaudan
26.3. SEVESO (parent company of ICMESA) required it for making the
At 12:37 p.m. on Saturday July 9, 1976, a bursting disc bacteriostatic agent hexachlorophene. It manufactured its
ruptured on a chemical reactor at the works of the own because the herbicide grades contained impurities
ICMESA Chemical Company at Meda near Seveso, unacceptable in this application. Between 1970 and 1976,
a town of about 17,000 inhabitants some 15 miles from some 370 te of the chemical were produced.
458 Lees’ Process Safety Essentials

The reaction was carried out in two stages. Stage 1 caused by chemicals. A mild case of chloracne usually
involved the alkaline hydrolysis of 1,2,4,5-tetrachloroben- clears within a year, but a severe case can last many
zene (TCB) using sodium hydroxide in the presence of a years. Other effects of TCDD include skin burns and
solvent ethylene glycol at a temperature of 170 2 180 C rashes and damage to liver, kidney, and urinary systems
to form sodium 2,4,5-trichlorophenate. The reaction mix- and to the nervous system. It appears to have an unusual
ture also contained xylene, which was used to remove the ability to interfere with the metabolic processes. There are
water by azeotropic distillation. In Stage 2, the sodium varying degrees of evidence for carcinogenic, mutagenic,
trichlorophenate was acidified with hydrochloric acid to and teratogenic properties. These are reviewed in the report.
TCP and purified by distillation. TCDD is a stable solid which is almost insoluble in
On completion of the Stage 1 reaction, some 50% of water and resistant to destruction by incineration except
the ethylene glycol would be distilled off and the temper- at very high temperatures.
ature of the reaction mixture lowered to 50 2 60 C by the
addition of water.
The process was a modification by Givaudan of a pro-
26.3.4 Events at Seveso
cess widely used in the industry. The conventional pro- The start of the batch began at 16:00 on Friday, July 9
cess used methanol rather than ethylene glycol and (Table 26.4). The reactor was charged with 2000 kg TCB,
operated at some 20 bar pressure. 1050 kg sodium hydroxide, 3300 kg ethylene glycol, and
In this reaction, the formation of small quantities of 600 kg xylene.
TCDD as a byproduct is unavoidable. At a reaction After the reaction had taken place, part of the ethyl-
temperature below 180 C, the amount formed would be ene glycol was distilled off, but the fraction removed
unlikely to exceed 1 ppm of TCP, but with prolonged was only 15% instead of the usual 50% so that most of
heating in the temperature range 230 260 C, it could the solvent was left in the vessel. Distillation was inter-
increase a thousand fold. rupted at 5:00 on July 10 and heating was discontinued,
During manufacture, nearly all (99.7%) of the TCDD but water was not added to cool the reaction mass.
formed concentrated in the distillation residues from The reactor was not brought down to the 50 2 60 C
which it was collected and incinerated. Only 0.3% found temperature range specified. The temperature recorder
its way into the TCP, giving a maximum concentration was switched off, with 158 C being the last temperature
of 10 ppb. recorded.
The reactor was a 138751 vessel with an agitator and The shift ended at 6:00. This time coincided with the
with a steam jacket supplied with steam at 12 bar. The closure of the plant for the weekend. The reactor was left
saturation temperature of steam at this pressure is 188 C. with the agitation turned off but without any action to
The controls on the reactor were relatively primitive. reduce the temperature of the charge.
There was no automatic control of the heating. During the weekend, with the steam turbine on reduced
The reactor was provided with a bursting disc set at load, the steam supply to the reactor jacket became super-
3.5 bar and venting direct to atmosphere. The prime heated at a temperature of about 300 C.
purpose of this disc was to prevent overpressure when
compressed air was being used on the reactor.
The works had an incinerator for the destruction of
26.3.5 The Release 1
hazardous plant residues at temperatures of 800 1000 C. The release contaminated the vicinity of the plant with
TCDD. Figure 26.1 is a map of the area, showing the
zones later established, and Table 26.5 gives the concen-
26.3.3 TCDD and its Properties trations later measured in these zones.
The properties of TCDD are given in the Seveso Report The area of Zone A was 108 ha (1.08 km2) with
and in Dioxin, Toxicological and Chemical Aspects by concentrations of TCDD averaging 240 µg/m2 and rising
Cattabeni et al. (1978) and by Rice (1982). 2,3,7,8 to over 5000 µg/m2. The area of Zone B was 269 ha
Tetrachlorodibenzo-p-dioxin is also known as TCDD or (2.69 km2) with concentrations averaging 3 µg/m2 but
dioxin. It is generated by the elimination of two mole- rising to 43 µg/m2. In the Zone of Respect, Zone R, which
cules of HCl from 2,4,5 trichlorophenol. had an area of 1430 ha (14.3 km2), the concentrations
TCDD is one of the most toxic substances known. The varied from indeterminable to 5 µg/m2.
lowest LD50 quoted is for guinea pigs and is 0.6 µg/kg,
which is a dose of 0.6 3 1029 per unit of body weight.
TCDD can be taken into the body by ingestion,
26.3.6 The Release 2
inhalation, or skin contact. A leading symptom of TCDD It is known that above 230 C, such a reactive mixture will
poisoning is chloracne, which is an acne-like skin effect undergo an exothermic decomposition reaction. However,
Chapter | 26 Case Histories 459

TABLE 26.4 Timetable of Events at Seveso


July 09 Friday 16:00 Final Reactor Batch Started

10 Saturday 5:00 Final reactor batch interrupted


12:37 Bursting disc on reactor ruptures
Barni visits houses near plant to warn against eating garden produce. He asks carabinieri to repeat
warning but they refuse
11 Sunday Barni and Paoletti inform von Zwehl. They are unable to contact Ghetti or his deputy. They inform Rocca
and Malgrati. Von Zwehl informs Sambeth
12 Monday Rocca and Uberti visit ICMESA. Letter of von Zwehl to Health Officer
13 Tuesday Report of Uberti to Rocca and Malgrati
14 Wednesday Sambeth and Vaterlaus inspect Dept B and contaminated area. Uberti writes to Provincial Health Officer
in Milan
15 Thursday Dubendorf Laboratories give first analyses showing high TCDD content. Sambeth telegraphs Waldvogel in
Turkey and asks von Zwehl to inform local authorities. Sambeth informs Hoffmann La Roche and seeks
clinical advice and permission to close plant. Rumours grow in Seveso. Citizens invade Uberti’s office.
Rocca and Uberti meet von Zwehl. They decide to declare polluted zone and to post warning notices

16 Friday Workforce goes on strike. Efforts are made to contact Ghetti in remote holiday farmhouse. Warning
notices are erected. Rocca, Uberti, and others meet von Zwehl again. Uberti insists on evacuation. Rocca
requests and obtains permission for evacuation from Deputy Prefect of Milan. Rocca contacts journalist
friend
17 Saturday II Giorno carries front-page headlines on ‘poison gas’ at Seveso. Press descends on town. Uberti contacts
Cavallaro
18 Sunday Cavallaro’s team of health inspectors investigate contaminated area. Rocca, Ghetti, Cavallaro, and
Adamo meet von Zwehl. Cavallaro presses to know identity of poison released. Adamo threatens to arrest
ICMESA management
19 Monday Waldvogel arrives in Seveso and offers local authority financial compensation, which is refused

20 Tuesday Cavallaro and Ghetti meet Vaterlaus in Zurich. Vaterlaus reveals that poison released was TCDD
Uberti’s letter to Provincial Health Officer in Milan arrives
16:00 Ghetti telephones Rocca to say substance was TCDD. Prefect in Milan is informed
21Wednesday 9:30 Meeting at prefecture. Rivolta and Carreri take over responsibility. Carabinieri arrest von Zwehl
and Paoletti
23 Friday 9:00 Meeting at prefecture. Rivolta decides against evacuation. Reggiani visits Rocca and urges
evacuation. Reggiani asked to leave Deputy Prefect’s meeting unheard
14:00 Rivolta reassures meeting in Seveso town hall. Reggiani argues with Rivolta and calls for
evacuation. Carabinieri try unsuccessfully to arrest Reggiani, who returns to Switzerland
24 Saturday 9:30 Meeting of regional health council. Vaterlaus presents map of locations affected. Council decides to
recommend evacuation and defines evacuation zone (part of Zone A)
26 Monday 179 people evacuated
29 Thursday Zone A extended and further 550 evacuated

accidents have been known to occur involving a reaction Investigations carried out after the accidents using
runaway above this temperature. differential thermal analysis showed that there exist two
Due to the interruption of the batch, the usual sharp slow exotherms (Theofanous, 1981, 1983). One starts at
reduction in the temperature of the charge at the termina- about 185 C, peaking at 235 C and giving a 57 C adiabatic
tion of the reaction did not take place, and after 7.5 h, the temperature rise. The other starts at about 255 C, peaking
explosion occurred. at 265 C and giving an estimated 114 C temperature rise.
460 Lees’ Process Safety Essentials

M r
The adiabatic induction times for the two exotherms are
e
ot
or Riv 2.1 and 0.5 h, respectively. The decomposition exotherm
wa R
y starts at about 280 2 290 C and shows a rapid pressure rise
at about 300 C.
Icmesa R
A mechanism for the reaction runaway has been
works Seves proposed by Theofanous. It states that due to layering of
lwa
y the reaction mix, the amount of residual heat in the upper
Rai
A section of the reactor wall was sufficient to raise the
Rive

temperature of the top layer of liquid to 200 220 C,


r

a temperature high enough to initiate exotherms leading


R to decomposition.
In conducting the final batch, there were failures of
adherence to operating procedures. Howard (1985)
B instances (1) failure to distil off 50% of the glycol;
R
(2) failure to add water; (3) failure to continue agitation;
(4) switching off of the reactor temperature recorder;
and (5) failure to bring the reactor temperature down
from its value of 158 C to the normal value of 50 60 C.
Riv

26.3.7 The Later Aftermath, Contamination,


er

R
and Decontamination
Mo

There was a prolonged period in which the authorities


tor
wa

tried to assess the situation and determine measures for


y

Milano dealing with it. They were advised by a team from


0 1 2 c. 30 km
Cremer and Warner (C&W), and an account has been
Kilometres given by Rice (1982).
(deduced scale) The release was modeled using fluid jet and gas disper-
sion models, and predictions were made on the probable
FIGURE 26.1 Plan of the Seveso area, showing Zones A and B and
Zone of Respect. Source: After Orsini (1980). ground level concentrations of TCDD. Accounts have been
given by Comer (1977) and Rice (1982). In this work,
the reaction mass was taken as 2800 kg ethylene glycol,
2030 kg trichlorophenol 2030, 542 kg sodium chloride, and
562 kg sodium hydroxide.
TABLE 26.5 Concentrations of TCDD in Zones A and B Estimates of the amount of TCDD generated in the
at Seveso reactor and dispersed over the countryside vary. Cattabeni
et al. (1978) give an estimated range of 0.45 3 kg
Zone Concentration (mg/m2)
released. The amount assumed in the C&W modeling
Mean Maximum Minimum work was 2 kg.
A1 580.4 5477 n.v.d Other parameters used in the modeling were bursting
disc rupture pressure at 376 kPa; a vent pipe diameter of
A2 521.1 1700 6.1
127 mm; and a discharge height above ground of 8 m.
A3 (north) 453.0 2015 1.7 The bursting disc rupture was assumed to a first approxi-
A3 (south) 93.0 441 n.v.d. mation to be due to the vapor pressure of ethylene glycol
at 250 C. From these data, the value obtained for the
A4 139.9 902 n.v.d.
vapor exit velocity was 274 m/s. Two limiting cases were
A5 62.8 427 n.v.d. considered for the discharge: Case A, pure vapor (density
A6 29.9 270 n.v.d. 1.61 kg/m3) and Case B, a two-phase vapor liquid mix-
ture (density 8.99 kg/m3). Eyewitness accounts indicated
A7 15.5 91.7 n.v.d.
that the actual event was intermediate between these two
B 3 43.8 n.v.d. extremes. For Case A, the total plume rise was estimated
at 83 m with the downwind distance to the maximum
n.v.d. 5 no value determined.
Source: After Orsini (1980). height as 103 m, and the corresponding figures for Case
B were 55 and 95 m, respectively.
Chapter | 26 Case Histories 461

The modeling work proved useful in defining the The Italian government set up the Commission of
problem and planning the decontamination. The results Inquiry to investigate the cause of the disaster and to
showed that beyond about 1.5 km, there was little differ- make recommendations. It also gave strong support to the
ence in the concentration estimates for the two cases. development of the EC Directive.
There was much agreement between the predicted and the
measured concentrations. This gave some confidence that 26.3.8.2 Siting of Major Hazard Installations
the amount released was indeed about 2 kg. There had
been some speculation by scientists that an amount as The release at Seveso affected the public because in the
high as 130 kg had been released. period since the site was first occupied, housing develop-
Various methods of decontamination were put forward ment had encroached on the area around the plant. The
by various parties, but were rejected by various reasons. accident underlined the need for separation between
TCDD is virtually insoluble in water, which reduces the hazards and the public.
threat to the water supply, but means that the effect of
rain is to transfer it from the vegetation to the soil rather 26.3.8.3 Acquisition of Companies Operating
than to remove it. Hazardous Processes
The decontamination measures actually carried out are
A lesson which has received relatively little attention is
described by Rice. They include collection of vegetation
the problem faced by a company which becomes owner
and cleaning of buildings by high intensity vacuum clean-
by acquisition of another company operating a hazardous
ing followed by washing with high-pressure water jets.
process. At Seveso, the problem was compounded by the
It is worth mentioning that people may also have been
fact that ICMESA was owned by Givaudan and the latter
affected by materials other than TCDD in the reactor
by Hoffmann La Roche, so that the company ultimately
charge. Marshall (1980) indicates that on the basis of the
responsible was twice removed. As a result, the directors
ground concentrations found in the contaminated area,
of the latter were not familiar with the hazards.
a minimum of 0.25 kg was released. He points out that
there were some 5 10 te of charge in the reactor and that
several tons of material, including sodium hydroxide, 26.3.8.4 Hazard of Ultra Toxic Substances
must have been ejected. He suggests that burns around Seveso threw into sharp relief the hazard of ultra toxic
the hands and other parts of the body suffered by 413 per- substances. The toxicity of TCDD is closer to that of a
sons who did not develop chloracne were probably caused chemical warfare agent than to that of the typical toxic
by sodium hydroxide. substance which the chemical industry is used to handling.
Data from the official report on ground concentrations As it happened, the British Advisory Committee on
in zones A and B are given in Table 26.5. Major Hazards was presenting its First Report, giving a
Kimbrough et al. (2010) provides a short account of scheme for the notification of hazardous installations,
the extent to which humans actually were exposed to the when Seveso occurred. It contained no notification propo-
chemicals from soil, food, or air. The large air borne sals for ultra toxics. In the Second Report, this deficiency
release of TCDD may have provided an acute inhalation was rectified. The EC Directive places great emphasis on
dose compared to the one ingested from locally grown toxic and ultra toxic materials.
produce after the release. Research on the toxicity of dioxins continues to shed
light on the effects that were observed after the incident.
Population-based studies (Pesatori et al., 2008) and short/
26.3.8 Some Lessons from Seveso long term morbidity and mortality (Pesatori et al., 2003)
Some of the numerous lessons to be learnt from the have been investigated. During the 20 year period,
Seveso disaster, many of which were brought out in the mortality and morbidity findings indicate that there is an
Commission’s report, are considered. increased risk from lymphoemopoietic neoplasm, digestive
system cancer, and respiratory system cancer.

26.3.8.1 Public Control of Major Hazard 26.3.8.5 Hazard of Undetected Exotherms


Installations
The characteristics of the reaction used had been investi-
The Seveso disaster had the effect of raising the general gated, and the company believed it had sufficient informa-
level of awareness of the hazards from chemical plants tion. It was well aware of the hazard from the principal
in Italy and in Europe, and highlighted the deficiencies in exotherm. Subsequent studies showed, however, that other,
the existing arrangements for the control of major weaker exotherms existed which, given time, could also
hazards. cause a runaway.
462 Lees’ Process Safety Essentials

The complete identification of the characteristics of 26.3.8.10 Planning for Emergencies


the reaction being operated is particularly important.
As the account given above indicates, the handling of the
If these are not known, or only partially known, there is
emergency was a disaster in its own right. Information on
always the danger that the design will not be as inherently
the chemical released and its hazards was not immedi-
safe as intended, that operating modifications will have
ately available from the company. There was failure of
unforeseen results, and/or that protective measures will be
communication between the company and the local
inadequate.
and regulatory authorities, and within those authorities.
After the incident, studies were carried out to better
Consequently, there was lack of action and failure to pro-
understand and investigate the trichlorophenol synthesis,
tect and communicate with the public. These deficiencies
such as the research conducted by Braun et al. (1999).
might in large part have been overcome by emergency
planning.
26.3.8.6 Hazard of Prolonged Holding
of Reaction Mass 26.3.8.11 Difficulties of Decontamination
The interruption of the reaction and the holding of the The incident illustrates the difficulties of decontamination
reaction mass for a prolonged period after the main reac- of land where the contaminant is both ultra toxic and
tion was complete without reducing the temperature gave insoluble in water.
time for the weak exotherms to occur and lead to
runaway.
26.4. MEXICO CITY
At about 5:35 a.m. on the morning of November 19,
26.3.8.7 Inherently Safer Design of 1984, a major fire and a series of explosions occurred at
Chemical Processes the terminal at San Juan Ixhuatepec (commonly known
The reactor was intended to be inherently safe to the as San Juanico), Mexico City. As a consequence of this
extent that the use of steam at 12 bar set a temperature incident, more than 500 people were killed, around 7000
limit of 188 C so that the contents could not be heated sustained injuries, and the terminal was destroyed.
above this by the heating medium. Unfortunately, this The accident was investigated by a team from TNO
feature was defeated by allowing the reaction charge to who visited the site about 2 weeks after the disaster and
stand hot for too long so that weak exotherms occurred, has published its findings (Pietersen, 1985, 1986a,b,
which the company did not know about. 1988). A further account has been given by Skandia
The bursting disc was not intended for reactor venting, International (1985).
but, as the Commission pointed out, if the set pressure of
the disc had been lower, the reactor would have vented at
26.4.1 The Site and the Plant
a lower temperature and with less dioxin in the charge.
The oldest part of the plant dated from 1961 to 1962, and
was thus over 20 years old. In the intervening period,
26.3.8.8 Control and Protection of Chemical residential development had crept up to the site. By 1984,
Reactors the housing was within 200 m of the installation, with
some houses within 130 m.
The control and protection system on the reactor was
The terminal was used for the distribution of LPG,
primitive. Operation of the reactor was largely manual.
which came by pipeline from three different refineries.
There was no automatic control of the cooling and no
The main LPG storage capacity of 16,000 m3 consisted of
high temperature trip.
6 spheres and 48 horizontal cylinders. The daily through-
The reactor was not designed to withstand a runaway
put was 5000 m3. The layout of the terminal with storage
reaction. It was not rated as a pressure vessel to withstand
tank capacities is shown in Figure 26.2. The two larger
pressure build-up prior to and during venting, the disc
storage spheres had individual capacities of 2400 m, and
was not designed for reaction relief, and there was no
the four smaller spheres had capacities of 1600 m3. The
tank to take the ultra toxic contents of the reactor.
site covered an area of 13,000 m2.
The plant was said to have been built to API stan-
dards, and much of it to have been manufactured in the
26.3.8.9 Adherence to Operating Procedures
United States.
The conduct of the final batch involved a series of failures A ground level flare was used to burn off excess gas
to adhere to the operating procedures. (see 7 in Figure 26.2). The flare was submerged in the
Chapter | 26 Case Histories 463

20
N
24

14
Unigas Gasomatico

1 2 6 9 19
5 13
3 22
8
7 4
10 12
15 18

11
16 17
23
21

Legenda:
7 Flare pit (2nd flare pit 16 LPG storage UNIGAS
1 2 Spheres of 2,400 m3, φ = 16.5 m see fig 1.3) 17 LPG storage GASOMATICO
2 4 Spheres of 1,600 m3, φ = 14.5 m 8 Pond 18 Bottling terminal
Pemex LPG installation 3 4 Cylinders of 270 m3, φ = 35.5 m, L = 32 m 9 Control room 19 Depot of cards with bottles
10 Pump house 20 Entrance
San Juan Ixhuatepec, 4 14 Cylinders of 180 m3, φ = 3.5 m, L = 21 m
11 Fire pumps 21 Rail car loading
Mexico City 5 21 Cylinders of 36 m3, φ = 2 m, L = 13 m 12 Road car loading 22 Store
6 6 Cylinders of 54 m3, φ = 2 m, L = 19 m 13 Gas bottle store 23 Water tank
3 Cylinders of 45 m3, φ = 2 m, L = 16 m 14 Pipe/valve manifold 24 Garrison
Scale 1:2,200 15 Water tower

FIGURE 26.2 Layout of the PEMEX site at Mexico City (Pietersen, 1985). Source: Courtesy of TNO.

ground to prevent the flame being extinguished by the cylinders had ruptured. The control room personnel tried
strong local winds. to identify the cause of the pressure fall but were not
Adjoining the PEMEX plant, there were distribution successful.
depots owned by other companies. The Unigas site was some The release of LPG continued for some 5 10 min.
100 200 m to the north and contained 67 tank trucks at the There was a slight wind of 0.4 m/s. The wind and the
time of the accident. Further away was the Gasomatico site, sloping terrain carried the gas toward the south-west.
with large numbers of domestic gas cylinders. People in nearby housing heard the noise of the escape
and smelled the gas.
When the gas cloud had grown to cover an area which
26.4.2 The Fire and Explosion 1 eyewitnesses put at 200 3 150 m with a height of 2 m,
Early on the morning of November 18, the plant was it found a flare and ignited at 5:40 a.m. The cloud caught
being filled from a refinery 400 km away through 8 in. fire over a large area, causing a high flame and a violent
feed pipeline. The previous day, the plant had become ground shock. When this general fire had subsided, there
almost empty, and refilling started during the afternoon. remained a ground fire, a flame at the rupture, and fires
The two larger spheres and the 48 cylindrical vessels had in some 10 houses.
been filled to 90% full, and the four smaller spheres to Workers at the plant now tried to deal with the escape.
about 50% full, so that the inventory on site was about One drove off to another depot to summon help. Five
11,000 m3, when the incident occurred. others who may have been on their way to the control
About 5:30 a.m., a fall in pressure was registered in room or to man fire pumps were found dead and badly
the control room and at a pipeline pumping station 40 km burned. At a late stage, someone evidently pressed the
away. The 8 in. pipe between sphere F4 and the Series G emergency shut-down button.
464 Lees’ Process Safety Essentials

footnote indicates, it is suggested by Skandia (1985) that


TABLE 26.6 Timetable of Events at Mexico City the initial explosion, or violent deflagration, was probably
A Seismograph Readings not recorded.
Numerous missiles were generated by the bursting of
1 5h 44 min 52 s 6 6h 49 min 38 s the vessels. Many of these were large and travelled far.
2 5h 46 min 01 s 7 6h 54 min 29 s Twenty-five large fragments from the four smaller spheres
3 6h 15 min 53 s 8 6h 59 min 22 s
weighing 10 2 40 te were found 100 2 890 m away.
Fifteen of the 48 cylindrical vessels weighing 20 te became
4 6h 31 min 5s 9 7h 01 min 27 s missiles and rocketed over 100 m, one travelling 1200 m.
5 6h 47 min 56 s Four cylinders were not found at all. The missiles caused
damage both by impact and by their temperature, which
B General Timetable
was high enough to set houses on fire.
5:30 Rupture of 8 in. pipe. Fall of pressure in control A timetable of events during the disaster is given in
room Table 26.6, Section B.
5:40 Ignition of gas cloud. Violent combustion and
high flame
5:45 First explosion on seismograph, a BLEVE 26.4.3 The Emergency
Fire department called Accounts of the emergency give little information about
the response of the on-site management in the emergency,
5:46 Second BLEVE, one of most violent
and deal mainly with the rescue and firefighting.
6:00 Police alerted and civilian traffic stopped The site became the scene of a major rescue operation
6:30 Traffic chaos which reached a climax in the period 8:00 2 10:00 a.m.
Some 4000 people participated in rescue and medical
7:01 Last explosion on seismograph, a BLEVE
activities, including 985 medics, 1780 paramedics, and
7:30 Continuing tank explosionsa 1332 volunteers. At one point, there were some 3000 people
11:00 Last tank explosion in the area. There were 363 ambulances and five helicopters
involved.
8:00 10:00 Rescue work at its height
The rescuers were at risk from a large BLEVE. The
12:00 18:00 Rescue work continues Skandia report states, ‘If a BLEVE had occurred during
23:00 Flames extinguished on last large sphere the later morning, a large number of those 3000 people
who were engaged in rescue and guarding would have
Disturbances 2 and 7 were the most intense with a Richter scale intensity
of 5. Skandia suggests that the first violent combustion may not have been killed’.
been recorded.
a
The fire services were called by surrounding plants and
Explosions of cylindrical vessels.
by individual members of the public at about 5:45 a.m.
They went into the plant area only 3 h after the start of the
incident. Initially, they moved toward the Gasomatico site
In the neighboring houses, some people rushed out
where a sphere fragment had landed and started a fire that
into the street, but most stayed indoors. Many thought it
caused the domestic gas cylinders to explode.
was an earthquake.
The fire brigade also fought the fire on the two larger
At 5:45 a.m., the first BLEVE occurred. About a min-
spheres, which had not exploded. They were at an appre-
ute later, another explosion occurred, one of the two most
ciable risk from a BLEVE in these two spheres. In this
violent during the whole incident. One or two of the
event, however, these burned themselves out. The last
smaller spherical BLEVEs produced a fireball 300 m in
flames on the spheres went out at about 11:00 p.m. Some
diameter.
200 firemen were at the site.
A rain of LPG droplets fell on the area. Surfaces
covered in the liquid were set on fire by the heat from the
fireballs. People also caught on fire.
There followed a series of explosions as vessels
26.4.4 The Fire and Explosion 2 2
suffered BLEVEs. There were some 15 explosions over a The TNO report gives technical information on the course
period of an hour and a half. BLEVEs occurred in the four of the disaster and on the fire and explosion phenomena
smaller spheres and many of the cylindrical vessels. which occurred during it.
The explosions during the incident were recorded on a The report discusses the effects of explosions, including
seismograph at the University of Mexico. The timing of vapor cloud explosions (VCEs), BLEVEs, and physical
the readings is given in Table 26.6 Section A. As the explosions; the effects of fire engulfment and heat
Chapter | 26 Case Histories 465

radiation; and the effects of missiles, including fragments Another risk was due to the large number of rescuers
from bullet tanks and spheres. who came on site due to the possibility of a BLEVE of
The report gives estimates of the overpressure from one of the larger spheres.
the BLEVEs of the principal vessels. It states that the
degree of blast damage to the housing was not great, that
the VCE effects were not responsible for major damage,
26.4.5.5 Fire Fighting in BLEVE Hazard
that the second explosion, a BLEVE, was the most violent Situations
and did damage houses, that the worst explosion damage The fire services appear to have taken a considerable risk
was probably from gases which had accumulated in in trying to fight the fire on the two larger spheres. The
houses, and that much of the damage was caused by fire. potential death toll if a BLEVE had occurred was high.
Films were available for many of the BLEVEs, though
not for the second, violent explosion. From this evidence,
the BLEVEs had diameters of 200 2 300 m and durations 26.4.5.6 Boiling Liquid Expanding
of some 20 s. Heavy direct fire damage was found at Vapor Explosions
distances up to about 300 m, which agrees reasonably After Flixborough the problem of vapor cloud explosions
well with the estimates of fireball size. received much attention. Mexico City demonstrates that
A very large fire burned on the site for about an hour boiling liquid expanding vapor explosions are an equally
and a half, punctuated by BLEVEs. Details are given on important hazard.
the number, size, and range of fragments from spheres The Mexico City incident represents the largest series
and bullet tanks. of major BLEVEs which have occurred, and provides
much information on BLEVEs. However, the TNO report
26.4.5 Some Lessons of Mexico City remains as the major source of information about this
accident. More than two decades have passed after the
Mexico City incident is one of largest incidents in process explosion at the San Juanico LPG terminal, and no offi-
safety and involving catastrophic BLEVEs. Things to be cial investigation report has been released to the public
learned from the incident may be as described here. by PEMEX or Mexican authorities.

26.4.5.1 Siting of Major Hazard Installations


26.5. BHOPAL
The high death toll at Mexico City occurred because
housing was too near the plant. At the time the plant was Early in the morning of December 3, 1984, a relief valve
constructed the area was undeveloped, but over the years lifted on a storage tank containing highly toxic methyl
the built-up area had gradually crept closer to the site. isocyanate (MIC) at the Union Carbide India Ltd (UCIL)
works at Bhopal, India. A cloud of MIC gas was released
26.4.5.2 Layout and Protection onto housing, including shantytowns, adjoining the site.
of Large LPG Storages Close to 2000 people died within a short period and tens
of thousands were injured.
The total destruction of the facility occurred because The accident at Bhopal is by far the worst disaster
there was a failure in the overall system of protection, which has ever occurred in the chemical industry. Its
including layout, emergency isolation, and water spray impact has been felt worldwide, but particularly in India
systems. and the United States. There are a lot of lessons we
should learn.
26.4.5.3 Gas Detection and Emergency
Isolation
26.5.1 The Site and the Works
One feature which might have averted the disaster is
more effective gas detection and emergency isolation. The works was in a heavily populated area. Much of the
The plant had no gas detector system and, probably as a housing development closest to the works had occurred
consequence, emergency isolation happened too late. since the site began operations in 1969, including the
growth of the J.P. Nagar shantytown. Although these
settlements were originally illegal, in 1984 the govern-
26.4.5.4 Planning for Emergencies
ment gave the squatters rights of ownership on the land to
One particularly unsatisfactory aspect of the emergency avoid having to evict them. Other residential areas which
was the traffic chaos which built up as residents sought to were affected by the gas cloud had been inhabited for
flee the area and the emergency services tried to get in. over 100 years.
466 Lees’ Process Safety Essentials

26.5.2 The Process and the Plant On each storage tank, there was a pressure controller
which controlled the pressure in the tank by manipulating
Monomethylamine (MMA) is reacted with excess phos- two diaphragm motor valves (DMVs), a make-up valve to
gene in the vapor phase to produce methylcarbamoyl admit nitrogen and a blowdown valve to vent vapor. Each
chloride (MCC) and hydrogen chloride, and the reaction tank had a safety relief valve (SRV) protected by a burst-
products are quenched in chloroform. The unreacted ing disc. They also had high temperature alarms and low
phosgene is separated by distillation from the quench and high level alarms.
liquid and recycled to the reactor. The liquid from the A vent gas scrubber (VGS) and a flare were provided
still is fed to the pyrolysis section where MIC is formed. to handle vented gases. The VGS was a packed column
The stream from the pyrolyser condenser passes as feed 5 ft 6 in. diameter in which the vent gases were scrubbed
to the MIC refining still (MRS). MIC is obtained as the with caustic soda. There were two vent headers going
top product from the still. The MIC is then run to storage. into the column: the process vent header (PVH), which
The MIC storage system (MSS) consisted of three collected the MIC system vents, and the relief valve vent
storage tanks, two for normal use and one for emergency header (RVVH), which collected the safety valve dis-
use. The tanks were 8 ft diameter by 40 ft long with a charges. Each vent header was connected both to the
nominal capacity of 15,000 USgal. They were made out VGS and the flare and could be routed to either. The vent
of 304 stainless steel with a design pressure of 40 psig at stack after the VGS was 100 ft (33 m) high.
121 C and with a hydrostatic test pressure of 60 psig. The VGS had the function of handling process vents
A diagram of the storage tank system is shown in from the PVH and of receiving contaminated MIC, in either
Figure 26.3. vapor or liquid form, and destroying it in a controlled
A 30 ton refrigeration system was provided to keep manner.
the tank contents at 0 C by circulating the liquid through The function of the flare was to handle vent gases
an external heat exchanger. from the carbon monoxide unit and the MMA vaporizer

FIGURE 26.3 Flow diagram of Tank 610 system (Bhushan and Subramanian, 1985) Source: Courtesy of Business India.
Chapter | 26 Case Histories 467

safety valve and also vent gas from the MIC storage factory, safety measures deficient’. These posters were
tanks, the MRS and the VGS. also distributed in the community.
In the 2 years preceding, the number of personnel on About a year before the accident, a ‘jumper line’ was
site were reduced, 300 temporary workers were laid connected between the PVH and the RVVH. Figure 26.3
off and 150 permanent workers pooled and assigned as shows the MIC storage tank and pipework arrangements.
needed to jobs, some of which they said they felt unquali- The jumper line is between valves 1 and 2. The object of
fied to do when interviewed. The production team at the the modification was to allow gas to be routed to the
MIC facility was cut from 12 to 6. VGS if repairs had to be made to one of the vent headers.
In June 1984, the 30 ton refrigeration unit cooling the
MIC storage tanks was shut down. The charge of Freon
26.5.3 MIC and its Properties refrigerant was drained from the system.
MIC is a colorless liquid with a normal boiling point In October, the VGS was turned off, apparently
of 39 C. It has a low solubility in water. It is relatively because it was thought unnecessary when MIC was only
stable when dry, but is highly reactive and in particular being stored not manufactured. In the same month, the
can polymerize and will react with water. It is flamma- flare tower was taken out of service, a section of corroded
ble and has a flashpoint of 218 C and a lower flamma- pipe leading to it was removed so that it could be
bility limit of 6% v/v. It is biologically active and highly replaced.
toxic. Another difficulty was experienced in pressurizing MIC
The high toxicity of MIC is indicated by the fact that its storage Tank 610. It appeared that since nitrogen was pass-
TLV at the time was 0:02 ppm. This is very low, relative to ing through the make-up valve satisfactorily, the blowdown
most typical compounds handled in industry. valve was leaking and preventing pressurization.
MIC is an irritant gas and can cause lung edema, According to plant workers, there were other instru-
but it also breaks down in the body to form cyanide. The mentation faults. The high temperature alarm had long
cyanide suppresses the cytochrome oxidase necessary for been faulty. There were also faults on the pressure con-
oxygenation of the cells and causes cellular asphyxiation. troller and the level indicator.
Information on the inhalation toxicity of MIC is given The plant had a toxic gas alarm system. This consisted
by Kimmerle and Eben (1964) and ten Berge (1985). of a loud siren to warn the public and a muted siren to
MIC can undergo exothermic polymerization to the warn the plant. These two sirens were linked and could
trimer, the reaction being catalyzed by hydrochloric acid be activated from a plant toxic alarm box. The loud siren
and inhibited by phosgene. It also reacts with water, iron could be stopped from the control room by delinking the
being a catalyst for this reaction. This reaction is strongly two. A procedure had been introduced according to which
exothermic. after delinking, the loud siren could be turned on only by
the plant superintendent.
Plant workers stated that on the morning of December 2,
26.5.4 Events Prior to the Release washing operations were undertaken. Orders were given to
In 1982, a UCC safety team visited the Bhopal plant. flush out the downstream sections of four filter pressure
Their report gave a generally favorable summary of the safety valves lines. These lines are shown in Figure 26.3.
visit, but listed ten safety concerns. In order to carry out this operation, Valve 16 on the diagram
Following this visit, valves on the MIC plant were was shut, Valves 18 2 21 and 22 2 25 were opened and
replaced, but degraded again. At the time of the accident, then Valve 17 was opened to admit water.
the instruments on Tank 610 had been malfunctioning for It was suggested water might have entered MIC stor-
over a year. age Tank 610 as a result of this operation—the water
Between 1981 and 1984, there were several serious washing theory. In this hypothesis, water evidently leaked
accidents on the plant. In December 1981, three workers through Valve 16, into the RVVH and passed through the
were gassed by phosgene and one died. Two weeks later, jumper line into the PVH and then into Tank 610. This
24 workers were overcome by another phosgene leak. would require that Valves 3 and 12 were open to connect
In February 1982, 18 people were affected by an MIC the tank to the PVH and Valves 1 and 2 open to connect
leak. In October 1982, three workers were injured and the RVVH to the PVH via the jumper line.
nearby residents affected by a leak of hydrochloric acid
and chloroform.
Following this latter accident, workers from the plant
26.5.5 The Release
posted a notice in Hindi which read, ‘Beware of fatal On the evening of December 2, a shift change took
accidents . . . lives of thousands of workers and citizens in place at the plant at 22:45. At 23:00, the control room
danger because of poison gas . . . Spurt of accidents in the operator noticed that the pressure in Tank 610 was
468 Lees’ Process Safety Essentials

10 psig. This was higher than normal but within the 26.5.6 Some Lessons Learned from
2 2 25 psig operating pressure of the tank. At the the Bhopal Incident
same time, the field operator reported a leak of MIC
near the VGS. At 00:15, the field operator reported an The Bhopal disaster was a wakeup call for the chemical
MIC release in the process area, and the control room industry around the world. Concerns from the public
operator saw that the pressure on Tank 610 was now before the incident were mainly health issues along with
at 30 psig and rising rapidly. He called the supervisor environmental losses. The Bhopal incident brought up
and ran outside to the tank. He heard rumbling sounds questions of public confidence in the chemical industries.
coming from the tank and a screeching noise from the There are numerous lessons learned from the Bhopal
safety valve and felt heat from the tank. He returned incident. Some of these lessons are now considered.
to the control room and turned the switch to activate
the VGS, but this was not in operational mode
because the circulating pump was not on. 26.5.6.1 Public Control of Major Hazard
At 00:20, the production supervisor informed the plant Installations
superintendent of the release. At 00:45 operations in the
The disaster at Bhopal received intense publicity for an
derivative unit were suspended due to the high concentra-
extended period and put major hazards on the public
tion of MIC.
agenda worldwide, but particularly in India and the United
At 01:00, an operator in this unit turned on the toxic
States, who had not reacted as strongly to Flixborough and
gas alarm siren. After 5 min the loud siren was switched
Seveso as Europe had. The incident was a stimulus
off leaving only the muted siren on.
to question the practices and attitudes toward the safety
At about the same time, the plant superintendent and
of chemical plants. Extensive legislative and industrial
control room operator verified that MIC was being emit-
improvements were brought up worldwide throughout
ted from the VGS stack into the atmosphere, and turned
decades.
on and directed water monitors at the stack fixed fire to
knock down the vapor.
Water was also directed at the MIC tank mound
and at the vent header for the VGS. Steam issued 26.5.6.2 Siting of and Development Control
from the cracks in the concrete showing that the tank at Major Hazard Installations
was hot. Very large numbers of people were at risk from the plant
One plant supervisor tried to climb the structure to at Bhopal. This situation was due in large part to the
plug the gas leak but was overcome, falling and breaking encroachment of the shantytowns, which came up to the
both legs. site boundary. Although these settlements were illegal,
Some time between 01:30 and 02:30, the safety valve the Indian authorities had acquiesced to them.
on Tank 610 reseated and the release of MIC ceased. In this instance, however, this was not the whole story.
At about 02:30, the loud siren was switched on again. The accident showed that the site was built close enough
The cloud of MIC gas spread from the plant toward to populated areas to present a hazard, when used for the
the populated areas to the south. There was a light wind production of a chemical as toxic as MIC. If the manufac-
and inversion conditions. ture of such a chemical was envisaged from the start, the
People in the housing around the plant felt the irritant problem may be regarded as one of siting. If not, it may
effect of the gas. Many ran out of their houses, some be viewed as one of intensification of the siting hazard.
toward the plant. Within a short period, animals and peo-
ple began to die.
At Railway Colony some 2 km from the plant,
26.5.6.3 Management of Major Hazard
where nearly 10,000 people lived, it was reported that
Installations
within 4 min 150 died, 200 were paralyzed, and 600
rendered unconscious, with 5000 people severely The plant at Bhopal was by any standards a major hazard
affected. and needed to be operated by a suitable competent man-
People tried to telephone the plant but were unable to agement. The standards of operation and maintenance do
get through. At 01:45, a magistrate contacted the plant not give confidence that this was so.
superintendent. There had been recent changes in the responsibility
The cloud of toxic gas hung around the area for the for the plant which suggest that the new management
entire day of December 3. During the day, it stopped may not have been familiar with the demands of a major
moving toward the city, but resumed its movement in that hazardous operation. However, many of the problems at
direction during the night. the plant appear to have predated these changes.
Chapter | 26 Case Histories 469

26.5.6.4 Highly Toxic Substances UCC’s West Virginia plant. UCIL had stated that it
regarded this inventory as undesirable, but was overruled
MIC is a highly toxic substance, much more toxic than
by the parent company, which wished to operate the same
substances such as chlorine which are routinely handled in
process at both plants.
the chemical industry. The hazard from such highly toxic
Processes are available for the manufacture of MIC
substances has perhaps been insufficiently appreciated.
which requires only small inventories of the material.
This hazard will only be realized if there is a mecha-
Moreover, carbaryl can be made by a route which does
nism for dispersion. At Bhopal, this mechanism was the
not involve MIC. The alternatives to the use of MIC are
occurrence of exothermic reactions in the storage tank.
discussed by Kletz (1988).

26.5.6.5 Runaway Reaction in Storage 26.5.6.10 Set Pressure of Relief Devices


The hazard of a runaway reaction in a chemical reactor is It is preferred from an operational viewpoint that the set
well understood, but such a reaction in a storage tank had pressure of a relief valve be such that the valve opens
received very little attention. At Bhopal, this occurred when a pressure rise threatens the integrity of the vessel,
due to ingress of water. When such a reaction could act but not when normal, minor operating pressure deviations
as the mechanism of dispersion for a large inventory of occur. However, when the cause of a potential pressure
a hazardous substance, the possibility of its occurrence rise is a runaway reaction, there is a penalty for a high set
should be carefully reviewed. pressure as it may allow the reaction to reach a higher
temperature and to proceed more rapidly before venting
26.5.6.6 Water Hazard in Plants starts. There is a need to balance these two factors.
In general terms, the hazard of water ingress into plants is
well known. In particular, water may contact hot oil and 26.5.6.11 Disabling of Protective Systems and
vaporize with explosive force or may cause a frothover, Alarms
it may corrode the equipment and it may cause a blockage
It was evident that the flare system was not viewed as a
by freezing. Bhopal illustrates the hazard of an exothermic
critical component for the protection of the plant, since it
reaction between a process fluid and water.
was allowed to remain out of commission for the 3 months
prior to the accident. The community alarm, which was
26.5.6.7 Relative Hazard of Materials activated 5 min after the leak, was then deactivated for
in Process and in Storage nearly an hour. It is essential that there be strict procedures
There has been a tendency to argue that the risks from for the disabling of any item which is critical for protection
materials in storage are less than from materials in process, and alarm systems. It is also very important that the time
since, although usually the inventories in storage are larger, for which an item is out of action be kept to a minimum.
the probability of a release is much less. The release at
Bhopal was from a storage tank, albeit from one associated 26.5.6.12 Maintenance of Plant Equipment
with a process. and Instrumentation
The 1982 UCC safety team drew attention to the pro-
26.5.6.8 Relative Priority of Safety blems in the maintenance of the plant. The Union
and Production Carbide Report Union Carbide Corporation (1985) gives
The features which led to the accident have been several examples of poor maintenance of plant equipment
described above. As indicated, the Union Carbide Report and instrumentation, and the ICFTU-ICEF report gives
Union Carbide Corporation (1985) itself refers to a num- further details. Workers stated that leaking valves and
ber of these. malfunctioning instruments were common throughout the
The ICFTU-ICEF report states that at the time of the plant.
accident the plant was losing money and lists a number of Maintenance was very slow. Although the crucial
measures which had been taken, apparently to cut costs. safety system was not operational, no action was con-
These include the manning cuts and the cessation of ducted to fix the problem. The flare system, which was a
refrigeration. critical protective system, had been out of commission
for 3 months before the accident. The vent gas scrubber
that was supposed to neutralize MIC gas with caustic
26.5.6.9 Limitation of Inventory in the Plant
soda was not fully operational. Temperature and pres-
The hazard at Bhopal was the large inventory of highly sure gauges were not giving reliable information and
toxic MIC. The process was the same as that used at alarms failed to work. The refrigeration system to cool
470 Lees’ Process Safety Essentials

MIC to around 0 C was shut down. Maintenance to keep Within the works, defects revealed by the emergency
a certain level of safety in the process should never be include the hesitation about the use of the siren system
overlooked. and the lack of escape routes.
The preliminary condition for emergency planning to
26.5.6.13 Isolation Procedures for Maintenance protect the public outside the works is provision to the
authorities of full information about the hazards. This was
A particular deficiency in the maintenance procedures not done. In consequence, the people exposed did not
was the failure to properly isolate the section of the plant know what the siren meant or what action to take, the
being flushed out by positive isolation using a slip plate hospitals did not know what they might be called on to
or equivalent means. The fact that the water may not handle, and so on.
have entered in this way does not detract from this lesson. Likewise, the essential action in an actual emergency
is to inform the authorities what has happened and what
26.5.6.14 Control of Plant and Process the hazards are. On the morning of the accident, the hos-
Modifications pitals were in the dark about the nature and effects of the
toxic chemical whose victims they were trying to treat.
A principal hypothesis to explain the entry of water into
Legislation to frame the outline of chemical emer-
Tank 610 is that the water passed through the jumper
gency responses, covering specification on emergency
line. The installation of this jumper line was a plant modi-
planning, emergency release notification, hazardous
fication. In a chemical plant, even the smallest change
chemical storage reporting requirements, and toxic chemi-
must go through reviews to verify any potential hazards
cal release inventory, is required at the community level.
before restarting the process. Company procedures called
for plant modifications to be checked by the main office
engineers, but they were evidently disregarded. The man- 26.5.6.17 Hazard Evaluations
agement of change (MOC) program is meant to prevent The MIC process was running without fully functional
any unforeseen consequences that may result in disastrous safety systems. The refrigeration system was shut down
outcomes. and gauges and alarms were not reliable. The vent gas
Insufficient MOC in Bhopal was one of the reasons scrubber was not fully operative and the flare tower and
why several safety measures at the plant failed to operate water curtain system were not designed to prevent any
at the time of the incident. The spare tank which was unexpected release of MIC. If any proper hazard evalua-
designed to be empty for emergency situations was not tions had been done, the management level would ques-
empty, and other tanks were filled over the recommended tion the safety of running the process without the factors
level with MIC. A properly managed MOC program is a mentioned above.
crucial part of safety procedures in running a chemical Hazard evaluations help to identify any inherent haz-
plant. ard or potential critical scenario when running the pro-
cess. It is recommended to evaluate the hazards that may
26.5.6.15 Information for Authorities and Public arise from the product and communicate the results to the
plant workers as well.
UCIL failed to provide full information on the substances
on site to the authorities, emergency services, workers,
and members of the public exposed to the hazards. 26.5.7 The Worldwide Impact of
Budget cuts in the company resulted in insufficient train- the Bhopal Incident
ing for many plant personnel, causing a lack of general
safety awareness. Many workers interviewed said they The Bhopal incident was a tragedy that attracted
had had no information or training about the chemicals. widespread attention to the concept of process safety.
The training, although done within the lower administra- Although process safety had been generally slighted in
tive level, can never be neglected, as it can play a signifi- the chemical industry, this incident provided the momen-
cant part in performing safety measures. This is more tum to implement process safety into the industry’s
crucial where process safety management principles and standard practices.
measures are insufficiently provided. The US Congress passed the Clean Air Act Amendments
(CAAA) in 1990 after Bhopal and several other domestic
incidents. It included some major provisions toward
26.5.6.16 Planning for Emergencies
the chemical industry and enlisted the Occupational Safety
The response of the company and the authorities to the and Health Administration (OHSA) to enforce the Process
emergency suggests that there were no effective emer- Safety Management of Highly Hazardous Chemicals
gency plans, procedures, or actions in place. (29 CFR 1910.119). The legislation also directed the
Chapter | 26 Case Histories 471

Environmental Protection Agency (EPA) to establish 26.6.2 Events Prior to the Explosion
the Risk Management Program (40 CFR 68) to target the
safety of personnel and communities in case of an acciden- On the previous day work began to clear three of the six
tal release. settling legs on Reactor No. 6, which were plugged. The
Despite the complex political and geographical three legs were prepared by a company operator and were
background, the European Union has tried to set a common handed over to the specialist maintenance contractors,
regulatory frame work toward the requirements of the Fish Engineering.
European Commission and the United Nations Economic At 8:00 a.m. on Monday, October 23, work began on
Commission for Europe. Before the Bhopal incident, the the second of the three blocked legs, Leg No. 4. The iso-
Seveso incident in 1976 initiated the ‘Council Directive lation procedure was to close the DEMCO ball valve and
82/501/EEC on the major-accident hazards of certain disconnect the air lines to it.
industrial actives’ known as ‘Seveso I’ to prevent serious The maintenance team partially disassembled the leg
consequences to the environments and public. After the and were able to remove part of the plug, but part
Bhopal incident, the regulation was amended and the remained lodged in the pipe 12 18 in. below the ball
Directive 88/610/EEC ‘Seveso II’ related to the Control of valve. One of the teams was sent to the control room to
Major Accident Hazards (COMAH) came into effect. seek assistance. Shortly after, at 1:00 p.m., the release
Before the Bhopal incident, Indian regulations lacked occurred.
a sufficient system to protect the public from industry Although both industry practice and Phillips corporate
hazards. The Environment Protection Act of 1986 was safety procedures require isolation by means of a double
established and, along with the Indian Factories Act of block system or a blind flange, at local plant level a pro-
1948, the Air Act was amended to regulate the facilities cedure had been adopted which did not conform to this.
dealing with hazardous materials. The Hazardous Waste It was subsequently established that the DEMCO ball
Rules of 1989 and the Public Liability Insurance Act of valve was open at the time of the release. The air hoses to
1991 followed, and the Environment Protection Rules the valve had been cross-connected so that the supply
were amended in 1992. which should have closed it actually opened it. The hose
connectors for the ‘open’ and ‘close’ sides of the valve
were identical, thus allowing this cross-connection to be
26.6. PASADENA made. Although procedures laid down that the air hoses
Shortly after 1:00 p.m. on October 23, 1989, a release should not be connected during maintenance, there was
occurred in a polyethylene plant at the Phillips 66 no physical barrier to the making of such a connection.
Company’s chemical complex at Pasadena, near Houston, The ball valve had a lockout system, but it was inade-
Texas. A vapor cloud formed and ignited, giving rise to a quate to prevent the valve being inadvertently or inten-
massive vapor cloud explosion. There followed a series tionally opened during maintenance.
of further explosions and a fire. Twenty-two people on
the site were killed and one later died from injuries, mak-
ing a death toll of 23. The number injured is variously
26.6.3 The Explosion
given as 130 and 300. The mass of gas released was estimated as some
A report on the investigation of the accident has been 85,200 lb of a mixture of ethylene, isobutane, hexene, and
issued by OSHA (1990). Other accounts include those hydrogen, which escaped within seconds. The release
of Mahoney (1990), T. Richardson (1991), and Scott was observed by five eyewitnesses. A massive vapor
(1992). cloud formed and moved rapidly downwind.
Within 90 2 120 s, the vapor cloud found a source of
ignition. Possible ignition sources were a gas-fired cata-
26.6.1 The Site and the Plant lyst activator with an open flame; welding and cutting
The Phillips works was sited in the Houston Chemical operations; an operating forklift truck; electrical gear in
Complex along the Ship Channel, the location of a num- the control building and the finishing building; 11 vehi-
ber of process companies. cles parked near the polyethylene plant office; and a
The plant in which the release occurred was Plant small diesel crane, although this was not operating.
V, one of the two active polyethylene plants in the com- The TNT equivalent of the explosion was estimated in
plex. The plant operated at high pressure (700 psi) and the OSHA report as 2.4 tons. An alternative estimate
high temperature. The process involved the polymeriza- from seismograph records is 10 tons.
tion of ethylene in isobutane, the catalyst carrier. Particles There followed two other major explosions, one when
of polyethylene settled out and were removed from set- two 20,000 gallon isobutane storage tanks exploded and
tling legs. the other when another polyethylene plant reactor failed
472 Lees’ Process Safety Essentials

catastrophically, the timings being some 10 15 min and would otherwise have had to cross the area of the explo-
some 25 45 min, respectively, after the initial explosion. sion to reach safety.
One witness reported hearing 10 separate explosions over The media were quickly aware of the explosion, and
a 2 h period. within an hour, there were on-site 150 media personnel
Debris from the explosion was found 6 miles from from 40 different organizations.
the site. The financial loss in this accident is comparable with,
All 22 of those who died at the scene were within 250 ft and may exceed, that of the Piper Alpha disaster.
of the point of release and 15 of them were within 150 ft. Redmond (1990) has quoted a figure of $1400 million,
Injuries which occurred outside the site were mainly divided almost equally between property damage and
due to debris from the explosion. business interruption losses.
The explosion resulted in the destruction of two On the basis of a review of company reports and of
HOPE plants. the defects found during the investigation of the disaster,
OSHA issued a citation to the company for wilful viola-
tions of the ‘general duty’ clause. The citation covered
26.6.4 The Emergency and the Aftermath the lack of hazard analysis; plant layout and separation
People in the immediate area of the release began running distances; flammable gas detection; ignition sources;
away as soon as they realized that gas was escaping. building ventilation intakes; and the fire water system;
The alarm siren was activated, but the level of noise in the permit system; and isolation for maintenance.
the finishing building was such that there was a question
whether some employees there failed to hear it.
The immediate response to the emergency was pro- 26.6.5 Some Lessons of Pasadena
vided by the site fire brigade, which undertook rescue and Some of the lessons to be learned from Pasadena are
care of the injured and began fighting the fire. Twenty- described here:
three people were unaccounted for, but for an extended
period, the area of the explosion remained dangerous to
enter. 26.6.5.1 Management of Major Hazard
Severe difficulties were experienced in fighting the Installations
fires resulting from the explosion. There was no dedicated The OSHA report details numerous defects in the man-
fire water system, water for firefighting being drawn from agement of the installation. Some of these are described
the process water system. The latter suffered severe rup- below.
ture in the explosion so that water pressure was too low
for firefighting purposes. Fire hydrants were sheared off
by the blast. Fire water had to be brought by hose from 26.6.5.2 Hazard Assessment of Major
remote sources such as settling ponds, a cooling tower, a Hazard Installations
water treatment plant, and a water main on a neighboring According to the report, the company had made no use of
plant. These difficulties were compounded by failures hazard analysis or an equivalent method to identify and
of the fire pumps. The electrical cables supplying power to assess the hazards of the installation.
the regular fire pumps were damaged by the fire so that
these pumps were put out of action. Further, of the three
backup diesel fire pumps, one was down for maintenance 26.6.5.3 Plant Layout and Separation Distances
and one quickly ran out of fuel. Despite these problems, The report was critical of the separation distances in the
the fire was brought under control within some 10 h. plant in several respects. It stated that the separation dis-
The handling of the emergency was handicapped by tances between process equipment plants did not accord
the facts that the intended command center had been dam- with accepted engineering practice and did not allow time
aged and that telephone communications were disrupted. for personnel to leave the polyethylene plant safely during
Telephone lines were jammed for some hours following the initial vapor release; and that the separation distance
the accident. between the control room and the reactors was insuffi-
The emergency response was coordinated by the site cient to allow emergency shut-down procedures to be car-
chief fire officer and involved the local Channel Industries ried out.
Mutual Aid (CIMA) organization, a cooperative of some
106 members in the Houston area.
26.6.5.4 Location of Control Room
More than 100 people were evacuated from the
administration building across the Houston Ship Channel As just mentioned, the control room was too close to the
by the US Coast Guard and by Houston fireboats; they plant. In addition, the control room was not built up to
Chapter | 26 Case Histories 473

the expected explosion over protection necessary for their 26.6.5.13 Dependability of Fire Pumps
location. It was destroyed in the initial explosion.
The electrical cables to the regular fire pumps were not
laid underground and were therefore vulnerable to damage
26.6.5.5 Building Ventilation Intakes by explosion and fire. One of the back-up diesel pumps
The ventilation intakes of buildings close to or downwind had insufficient fuel and one had been taken out for main-
of the hydrocarbon processing plants were not arranged tenance without informing the chief fire officer.
so as to prevent intake of gas in the event of a release.
26.6.5.14 Audibility of Emergency Alarm
26.6.5.6 Minimization of Exposure of Personnel As described, the level of noise in some areas was such
Closely related to this, there was a failure to minimize the that the employees might not have been able to hear the
exposure of personnel. Not only the control room but the siren.
finishing building had relatively high occupancy.
26.6.5.15 Follow-Up of Audits
26.6.5.7 Escape and Escape Routes The OSHA report criticized the company’s failure to act
As already stated, the separation distances were not such upon reports issued previously by the company’s own
as to allow personnel on the polyethylene plant to escape safety personnel and by external consultants, which drew
safely. Further, the only escape route available to people attention to unsafe conditions.
in the administration block (other than across the ship
channel) was across the area of the explosion. 26.6.5.16 Planning for Emergencies
The disaster highlighted a number of features of emer-
26.6.5.8 Gas Detection System gency planning. The company had put a good deal of
Despite the fact that the plant had a large inventory of effort into planning and creating personal relationships
flammable materials held at high pressure and tempera- with the emergency services, by means such as joint exer-
ture, there was no fixed flammable gas detection system. cises, and these paid off. The value of planning, training,
and personal relations was one of the most positive
26.6.5.9 Control of Ignition Sources lessons drawn.
Another area in which a proactive approach proved
The control of ignition sources around the plant was beneficial was in relations with the media. Senior person-
another feature criticized in the OSHA report. nel made themselves available, and the company evidently
felt it received fair treatment.
26.6.5.10 Permit-to-Work Systems One weakness of the emergency planning identified
was that it had not envisaged a disaster of the scale which
The OSHA report stated that an effective permit system actually occurred.
was not enforced for the control of the maintenance activ- The incident brought out the need to be able to respond
ities either of the company’s employees or of contractors. clearly to calls from those liable to be affected about the
toxicity of the fumes and smoke generated in such an
26.6.5.11 Isolation Procedures for Maintenance event.
In this incident, the sole isolation was a ball valve, which The behavior of rescue helicopters posed a problem.
was meant to be closed but was in fact open. There was Personnel on the ground had no means of communication
no double block system or blind flange. with them and the craft tended to come in low, creating
The practice of not providing positive isolation was a the danger of blowing flames or toxic fume onto those
local one and violated corporate procedures. The implica- below. A need was identified for altitude and distance
tion is that it had not been brought to light by any safety guidelines for helicopters.
audits conducted.
26.7. CANVEY REPORTS
26.6.5.12 Integrity of Fire Water System
The most comprehensive hazard assessment of non-
The practice of relying for fire water on the process water nuclear installations in the United Kingdom is the Canvey
system and the failure to provide a dedicated fire water study, carried out for the HSE by SRD.
system meant that the fire water system was vulnerable to The first phase of the work is described in Canvey: An
an explosion. Investigation of Potential Hazards from Operations in the
474 Lees’ Process Safety Essentials

Canvey Island/Thurrock Area (the First Canvey Report) 5. To quantify the probability of the selected failures
(HSE, 1978). The report is in two parts: Part 1 is an intro- occurring and their consequences.
duction by the HSE and Part 2 is the SRD study.
The investigation involved the identification of the
The origin of the investigation was a proposal to
principal hazards of the installations and activities in the
withdraw planning permission for the construction of an
area, the assessment of the associated risks to society and
additional refinery in the area. Two oil companies,
to individuals, and the proposal of modifications intended
Occidental Refineries Ltd and United Refineries Ltd, had
to reduce these risks.
been granted planning permission for the construction of
Some 30 engineers were engaged in the investigation,
oil refineries. The construction of the Occidental refinery
which cost about d400,000.
was begun in 1972, but was halted in 1973 pending a
major design study review. United Refineries had valid
planning consents, but had not started construction. It was
a public inquiry into the possible revocation of the plan- 26.7.2 First Canvey Report: Installations
ning permission for the United Refineries development and Activities
which gave rise to the investigation.
Responses to the First Canvey Report centered mainly The principal hazardous installations and activities identi-
on two aspects: the methodology used and the magnitude fied in the investigation are summarized in Table 26.7.
of the assessed risks. The HSE commissioned further
work, leading to the Second Canvey Report (HSE, 1981),
in which the methodology used is revised and the 26.7.3 First Canvey Report: Identified
assessed risks are rather lower.
Hazards
The investigation identified several principal hazards in
the area. These are:
26.7.1 First Canvey Report
1. oil spillage over bund;
The terms of reference of the investigation were: 2. LNG vapor cloud release (1000 t);
3. LPG vapor cloud release (1000 t);
In the light of the proposal by United Refineries Limited to
4. ammonium nitrate explosion (4500 t of 92% solution);
construct an additional refinery on Canvey Island, to investigate
5. ammonia vapor cloud release (1000 t);
and determine the overall risks to health and safety arising from
6. hydrogen fluoride cloud release (1000 t).
any possible major interactions between existing or proposed
installations in the area, where significant quantities of danger- The figures in parentheses indicate standard cases
ous substances are manufactured, stored, handled, processed considered in the study.
and transported or used, including the loading and unloading of A severe fire might occur if there is an escape of
such substances to and from vessels moored at jetties; to assess flammable liquids from storage so that large quantities flow
the risk; and to report to the Commission. down into the residential area. This hazard is presented by
the large storages of flammable liquids.
The members of the investigating team were
A severe vapor cloud fire and/or explosion might
appointed as inspectors of the HSE under the provisions
occur if there is a spillage of LNG so that a vapor cloud
of the Health and Safety at Work etc. Act of 1974 and
forms and ignites. This hazard is presented by the large
were given specified powers to enable them to make the
LNG terminal and storage at British Gas. The spillage
necessary inquiries. The overall approach taken in the
might occur at sea or on land.
investigation was
Similarly, a severe vapor cloud fire and/or explosion
1. To identify any potentially hazardous materials, their might occur if there is a spillage of LPG so that a vapor
location and the quantities stored and in process. cloud forms and ignites.
2. To obtain and review the relevant material properties A severe explosion might occur if there is a rupture of
such as flammability and toxicity. an ammonium nitrate storage tank.
3. To identify the possible ways in which failure of A severe toxic release might occur if there is a spillage
plants might present a hazard to the community. of ammonia. The spillage might occur at sea or on land.
4. To identify possible routes leading to selected failures. A severe toxic release might also occur if there is a
Typically, the factors examined included operator rupture of storage or process plant containing hydrogen
errors, fatigue or aging of the plant, corrosion, loss of fluoride. This hazard is presented by the alkylation facili-
process control, overfilling, impurities, fire, explosion, ties at Shell, at the Mobil extension and at the proposed
missiles and flooding. Occidental refinery.
TABLE 26.7 First Canvey Report: Principal Hazardous Installations and Activities at Canvey

Location Company Installation or Activity Storage Employees Transport In Transport Out


Canvey British Gas LNG terminal Fully refrigerated storage of LNG (atmospheric pressure, 200 Sea Mainly pipeline (as vapor)
Island Corporation , 2162 C) 6 3 4000 t above-ground tanks 2 3 1000 t but some road
above ground tanks 4 3 20,000 t in ground tanks. Fully
refrigerated storage of butane (atmospheric pressure,
,10 C) 1 3 10,000 t tank 2 3 5000 t tanks
Texaco Ltd Petroleum products storage Atmospheric storage of petroleum products .80,000 t 130 Sea Pipeline, road, sea
total capacity
London and Coastal Flammable and toxic Atmospheric storage of liquids .300,000 t total capacity 50 Mainly sea but Pipeline (Texaco oil). Rest
Wharves Ltd liquids storage some road mainly road but some sea
Occidental Oil refinery (proposed) Pressure storage of LPG (atmospheric temperature)
Refineries Ltd 2 3 750 t propane spheres 2 3 400 t butane spheres
United Refineries Ltd Oil refinery (proposed) Pressure storage of LPG (atmospheric temperature)
4 3 200 t propane spheres 3 3 900 t butane spheres.
Process and storage containing hydrogen fluoride
Coryton Mobil Oil Co. Ltd Oil refinery Pressure storage of LPG (atmospheric temperature) 800 LPG produced Pipeline, road, rail, sea
1 3 1000 t 1 14 other vessels, giving 4000 t total capacity on site

Oil refinery (extension) Pressure storage of LPG (atmospheric temperature) LPG produced Pipeline, road, rail, sea
4 3 1000 t LPG spheres. Fully refrigerated storage of LPG on site
(atmospheric pressure) 1 3 5000 t tank. Process and
storage containing hydrogen fluoride
Calor Gas Ltd LPG terminal Pressure storage of LPG (atmospheric temperature) 100 Pipeline (from oil Road (cylinders, bulk
3 3 60 t propane vessels 2 3 60 t butane refineries) tankers)
vessels 1 cylinders, giving 500 t total inventory

Shell Haven Shell Oil U.K. Ltd Oil refinery Pressure storage of LPG (atmospheric temperature) 1900 LPG produced Pipeline, road, rail, sea
1 3 1700 t butane sphere 1 3 other butane spheres, on site
giving 3200; t total butane capacity 4 3 400 t propane
spheres 3 3 135 t LPG horizontal vessels
Fully refrigerated storage of liquid anhydrous ammonia Sea Sea, road, occasionally rail
(atmospheric pressure, 233 C) 1 3 14,000 t tank. Process
and storage containing hydrogen fluoride 2 3 40 t vessels

Stanford-le- Fisons Ltd Ammonium nitrate plant Storage of 92% aqueous ammonium nitrate solution 80 Ammonium nitrate Road
Hope 1 3 5000 t tank 1 3 2000 t tank produced on site

Ammonia storage Semi-refrigerated pressure storage of liquid anhydrous Rail Ammonia used on site
ammonia (pressure above atmospheric, 6 C) 1 3 2000 t
sphere
Canvey/ General Transport of hazardous
Thurrock area materials by river, road,
rail, and pipeline
476 Lees’ Process Safety Essentials

26.7.4 First Canvey Report: Failure The projects that were initiated as part of the investi-
and Event Data gation were:

The investigation required the estimation of the 1. consideration of known history of identified storage
probabilities of various occurrences and of their tanks and their possibility of failure;
consequences. 2. probability of particular storage tanks or process vessels
Some of the sources of information on such probabilities being hit by missiles caused by fires or explosions
used in the study were: on site or adjacent sites, by fragmentation of rotating
machines or pressure vessels, or transport accidents;
1. UK industries, including oil, chemical and other process 3. effects of vapor cloud explosions on people, houses,
industries and transport; engineering structures, etc.;
2. government organizations such as those concerned 4. evaporation of LNG from within a containment area
with fire, road, rail, sea, and air transport; on land or from a spill on water;
3. professional institutions, for example, Institution of 5. special problems of frozen earth storage tanks for
Chemical Engineers, Institution of Civil Engineers, LNG and the effect of flooding;
American Institute of Chemical Engineers; 6. study of possible failures in handling operations;
4. international safety conference proceedings, for exam- 7. consideration of the possible benefits and practicality
ple, loss prevention in the process industries, ammonia of evacuation;
plant safety, and hazardous materials spills; 8. civil engineering aspects of the sea-wall—the chance of
5. industry-based associations, for example, Chemical it being breached by subsidence, explosion or impact
Industries Association, Institute of Petroleum, American of ships, consideration of the timing of improved
Petroleum Institute, Liquefied Petroleum Gas Industry defenses, consideration of the time for floods to rise;
Technical Association; 9. statistics of ship collisions and their severity, ground-
6. international insurance interests, for example, Lloyds, ings, etc., applying extensive world experience to the
Det Norsk Veritas, and industrial risk insurers, Fire Canvey Island area;
Protection Association; 10. reliability and analysis of fluid handling practices,
7. overseas government and international agencies, ship to shore, and store to road vehicles and pipelines;
for example, US Coast Guard, US Department of 11. toxicology of identified hazardous substances;
Transportation, OECD, and EEC; 12. studies to determine the lethal ranges for various
8. specialized research laboratories; releases of toxic or explosive materials leading to a
9. individual subject specialists known or recommended number of special studies such as
to the investigating team. a. the behavior of ammonia spilt on water or land, and
The degree of uncertainty associated with the probability b. an assessment of the relative importance of explo-
estimates is indicated by the following code: sion or conflagration from a cloud of methane or
liquefied petroleum gas.
a. assessed statistically from historical data—this method
is analogous to the use of aggregate estimates in The subjects that are considered in appendices to the
economic forecasting; report are:
b. based on statistics as far as possible but with some 1. a review of current information on the causes and
missing figures supplied by judgment; effects of explosions of unconfined vapor clouds
c. estimated by comparison with previous cases for (F. Briscoe);
which fault tree assessments have been made; 2. fires in bunds—calculations of plume rise and position
d. ‘dummy’ figures—likely always to be uncertain, of downwind concentration maximum (R. Griffiths);
a subjective judgment must be made; 3. a quantitative study of factors tending to reduce the
e. not used; hazards from airborne toxic clouds (Q.R. Beattie);
f. fault tree synthesis, an analytically based figure which 4. the dispersal of ammonia vapor in the atmosphere with
can be independently arrived at by others. particular reference to the dependence on the conditions
of emission (F. Abbey, R.F. Griffiths, S.R. Haddock,
G.D. Kaiser, R.J. Williams, and B.C. Walker);
26.7.5 First Canvey Report: Hazard Models 5. statistics on fires and explosions at refineries
and Risk Estimates (Q.H. Bowen);
6. missiles—penetration capability (E.A. White);
The investigation involved the study of a wide range of 7. discussion of data base for pressure vessel failure
hazards and scenarios. rate (T.A. Smith);
Chapter | 26 Case Histories 477

8. risk of aircraft impacts on industrial installations in 14. dispersion of ammonia and hydrogen fluoride vapor
the vicinity of Canvey Island (L.S. Fryer); clouds;
9. the dispersion of gases that are denser than air, with 15. factors mitigating casualties from a toxic release;
LNG vapor as a particular example (G.D. Kaiser); 16. road tanker hazards;
10. not used; 17. evacuation.
11. the risk of a liquefied gas spill to the estuary (D.F.
For a more complete discussion of the methods and
Norsworthy);
of the background to and application, please consult the
12. the toxic and airborne dispersal characteristics of
report.
hydrogen fluoride (Q.R. Beattie, F. Abbey, S.R.
Haddock, G.D. Kaiser);
13. transient variation of the wall temperature of an LNG 26.7.6 First Canvey Report: Assessed
above-ground storage tank during exposure to an
LNG fire in an adjacent bund (I.R. Fothergill);
Risks and Actions
14. the escape of 1000 t of anhydrous ammonia from The hazards described in Section 26.7.3 were assessed
a pressurized storage tank (L.S. Fryer, G.D. Kaiser, using appropriate failure and event data as described in
and B.C. Walker); Section 26.7.4. Hazard models and risk estimates to be
15. graphical calculation of toxic ranges for a release of used are listed in Section 26.7.5.
1000 t of ammonia vapor (J.H. Bowen); The results of the risk assessments are presented by
16. effect of unbunded spill of hydrocarbon liquid from the investigators as risks of causing casualties, that is,
refinery at Canvey Island (A.N. Kinkead); severe hospitalized casualties or worse. This is in accor-
17. estimated risk of missile damage causing a vapor dance with established practice (e.g., Department of
cloud release from existing and proposed LPG stor- Defense, n.d.; Glasstone, 1964). It was considered mis-
age vessels at the Mobil refinery, Coryton (D.F. leading to attempt to distinguish between severe injury
Norsworthy); and death.
18. risks of accidents involving road tankers carrying These results have several interesting features. The
hazardous materials (L.S. Fryer); hazards may be ranked for societal risks in order of des-
19. toxicology of lead additives (S.R. Haddock); cending frequency for accidents of different magnitude.
20. compatibility of materials stored at London and The hazard arising from the very large quantities of
Coastal Wharves Ltd (S.R. Haddock); LNG stored is a serious one, but is no worse than that
21. blast loading on a spherical storage vessel (Q. Wall); from the considerably smaller quantities of LPG.
22. statistical comment on data on distribution of cracks The obvious hazards of LNG, LPG, and ammonia are
found on inspection of steel vessel (Q.C. Moore); equaled by others, such as oil and hydrogen fluoride,
23. calculation of resistance of ship hull to collision which are perhaps less well appreciated.
(A.N. Kinkead); The relative importance of the hazards changes with
24. reduction of apparent risk by shared experience the scale of the accident. For the smaller scale accidents,
(Q.H. Bowen). oil spillage, flammable vapor clouds, and toxic gas clouds
are all important. As the scale increases, it is the toxic gas
The treatment of some of the topics which is given in
clouds that dominate.
the report is now described as follows:
There are a number of interactions identified both
1. failure of pressure vessels; within sites and between sites. These include the threat to
2. failure of pressure piping; LPG storage at Mobil from the Calor Gas site, to the oil
3. failure of pipelines; storage at Texaco from explosives barges, to the ammonia
4. generation of and rupture by missiles; sphere at Fisons from rotating machinery, and from the
5. crash of and rupture by aircraft; ammonium nitrate plant at Fisons, to various installations
6. ship collision and other accidents; from process and jetty explosions, and possibly to the
7. flow of a large release of oil; ammonia storage tank at Shell from explosion in the
8. temperature of the wall of an LNG tank exposed to Shell refinery.
an LNG fire in an adjacent bund; The relative hazard of the pressure storage of anhy-
9. evaporation of LNG and ammonia on water; drous ammonia at Fisons is much greater than that of the
10. dispersion of an LNG vapor cloud; refrigerated storage of the same chemical at Shell. The
11. unconfined vapor cloud fire and explosion; risk for the latter was assessed as negligible with the
12. ammonium nitrate explosion; possible exception of rupture by an explosion.
13. toxicity of chlorine, ammonia, hydrogen fluoride, The assessed societal risks are shown in Figure 26.4.
and lead additives; Figure 26.4(a) gives the societal risks for all the existing
478 Lees’ Process Safety Essentials

32 FIGURE 26.4 First Canvey Report:


30 societal risks for existing installations
Probability of exceeding scale of accident indicated

and proposed developments (Health


28 Existing and Safety Executive, 1978). (a) All
26 Existing, after existing installations. (b) All proposed
24 improvements suggested developments. Source: Courtesy of
22 HM Stationery Office.
(chances in 10 000/y)

20
18
16
14
12
10
8
6
4
2
0
10 1000 5000 10 000 15 000 18 000
Scale of accident (no. of casualties)
(a)
Probability of exceeding scale of accident

20
indicated (chances in 10 000/y)

18 Proposed developments
16 Proposed developments,
after improvements suggested
14
12
10
8
6
4
2
0
10 1000 5000 10 000 15 000 18 000
Scale of accident (no. of casualties)
(b)

installations. It shows that the risk of an accident causing 2. LNG tank flooding—construction of a dike;
more than 10 casualties is 31 3 10 4/year. Figure 26.4(b) 3. spontaneous failure of LPG spheres—high standard
gives the societal risks for all the proposed developments. of inspection;
It shows that the risk of an accident causing more than 10 4. LPG vessel rupture by missile (Mobil)—measures
casualties is 16 3 10 4/year. including fitting of pressure relief valves on cylinders
The assessed individual risks are given for all the at Calor Gas depot;
existing installations as given in Table 26.8. 5. LPG tank failure at BG jetty—improvement of bund;
The investigators made a number of recommendations 6. LPG pipeline failure (BG)—removal of pipeline;
for the reduction of the hazards. These included 7. spontaneous failure of ammonia sphere—high stan-
dard of inspection, control of ammonia purity;
1. oil spillage—construction of a simple containing 8. ammonia release at Shell jetty—provision of water
wall around the London and Coastal Wharves and sprays at jetty;
Texaco sites and the proposed Occidental and UR 9. HF plant rupture (Mobil, Occidental)—provision of
refinery sites; water sprays;
Chapter | 26 Case Histories 479

TABLE 26.8 First Canvey Report: Rank Order of Societal Risks for Some Principal Hazards of Existing and Proposed
Installations at Canvey
Frequencies in Units of 10 6/year

.10 Casualties .4500 Casualties .18,000 Casualties

1 Oil spillage 1366 1 Ammonia vapor cloud 258 1 HF vapor cloud 100
2 LPG vapor cloud (BG) 970 2 HF vapor cloud 246 2 Ammonia vapor cloud 73
3 Ammonia vapor cloud 735 3 Oil spillage 150 3 LNG vapor cloud 7
4 LPG vapor cloud (others) 637 4 LPG vapor cloud (other) 96
5 LNG vapor cloud 497 55 LNG vapor cloud 80
6 HF vapor cloud 464 55 LPG vapor cloud (BG) 80

7 AN explosion 85 7 AN explosion 17

Source: After Health and Safety Executive, (1978).

TABLE 26.9 First Canvey Report: Rank Order of Societal Risks for Some Principal Hazards of Existing and Proposed
Installations at Canvey, After Improvements Suggested
Frequencies in Units of 10 6/year

. 10 Casualties .4500 Casualties .18,000 Casualties


1 LPG vapor cloud (other) 421 1 Ammonia vapor cloud 71 1 HF vapor cloud 24
2 LNG vapor cloud 396 2 HF vapor cloud 67 2 Ammonia vapor cloud 18
3 Ammonia vapor cloud 240 3 LNG vapor cloud 66 3 LNG vapor cloud 7

4 HF vapor cloud 115 4 LPG vapor cloud (other) 63


5 AN explosion 8 5 AN explosion 2

Source: After Health and Safety Executive (1978).

10. ship collision—strict enforcement of the speed as given in Table 26.9. Oil spillage and LPG vapor cloud
restriction of 8 knots; (BG) are eliminated.
11. road tanker hazards—road tanker traffic restriction to
new road only (if road built).
26.7.7 Second Canvey Report
The assessed effect of the proposed modifications is to
In 1981, the Second Canvey Report was issued, entitled
eliminate the hazards from oil spillage, from LPG vapor
Canvey: A Review of Potential Hazards from Operations in
cloud due to explosion at the jetty and pipeline failure
the Canvey Island/Thurrock Area Three Years after
at British Gas, and from vessel rupture at Mobil due to
Publication of the Canvey Report (HSE, 1981). The report
missiles from Calor Gas, and to reduce greatly many of
was again in two parts: Part 1, an introduction by the HSE
the other hazards.
and Part 2, the SRD study. In the next two sections,
The rank order for societal risk of the hazards assum- accounts are given of the reassessed risks and actions and
ing that the proposed modifications are carried out is then of some technical aspects of the second report.
480 Lees’ Process Safety Essentials

26.7.8 Second Canvey Report: Reassessed near Rozenburg, and major petrochemicals complexes
Risks and Actions operated by Shell, Gulf, and Esso.

The risks assessed in the Second Canvey Report are


generally less than those in the first report. Thus, for 26.8. RIJNMOND REPORT
example, at Stanford-le-Hope the individual fatality risk 26.8.1 The Investigation
is given as 0.6 3 10 4/year, as opposed to 5 3 10 4/year
in the first report, a reduction by a factor of 8. The aim of the work was to evaluate the methods of
The HSE in Part 1 adduce the following reasons for risk assessment for industrial installations and to obtain
this: (1) physical improvements already carried out; (2) experience in the practical application of these methods.
changes in operation; (3) detailed studies by companies; Such evaluation was considered to be essential before any
(4) changes in assessment techniques; (5) correction of decision could be made on the role of such methods in
errors; and (6) further changes firmly agreed but yet to be the formulation of safety policy.
made. The COVO steering committee was reinforced with
One major change in operation was the cessation of experts from industry. The overall approach taken in the
ammonia storage at Shell. investigation was:
In a number of cases in the first report, estimates of the 1. to collect basic data and define the boundary limits of
risks had to be made without benefit of detailed studies. the installation;
Subsequent studies by the companies involved showed that 2. to identify the potential failure scenarios;
in some cases the risks had been overestimated. A case in 3. to select and apply the best available calculation models
point was the limited ammonia spill at Fisons. for physical phenomena;
In a large proportion of cases, the effect of the use of 4. to collect and apply the best available basic data and
improved models for heavy gas dispersion was to reduce models to calculate the probabilities of such events;
the travel distance of the cloud and hence the risks. 5. to choose and develop different forms of presentation
The HSE describes a number of further actions taken of the final results;
or pending to reduce the risks. 6. to investigate the sensitivity of the results for variations
in the assumptions used and to estimate the accuracy
and reliability of these results;
26.7.9 Second Canvey Report: Technical 7. to investigate the influence of risk reducing measures.
Aspects
The Second Canvey Report revisits some of the frequency 26.8.2 Installations and Activities
estimates and gives a revised set of hazard models and The six installations studied were Acrylonitrile storage
injury relations, covering the following aspects: (1) emis- (Paktank); Ammonia storage (UKF); Chlorine storage
sion; (2) gas dispersion; (3) ignition; (4) fire events; (5) (AKZO); LNG storage (Gasunie); Propylene storage
vapor cloud explosions; and (6) toxic release. With these (Oxirane); Hydrodesulfurizer (Shell).
improvements, the second report may give more reason-
able assessment.
Another comprehensive hazard assessment is the 26.8.3 Event Data
Rijnmond study carried out for the Rijnmond Public The event data used in the study were a mixture of
Authority by Cremer and Warner. generic data, plant data, and estimates.
This work is described in Risk Analysis of Six
Potentially Hazardous Industrial Objects in the Rijnmond
Area, a Pilot Study issued by the Rijnmond Public 26.8.4 Hazard Models
Authority (1982) (the Rijnmond Report). The principal hazard models used are given in Table 26.10.
The Rijnmond is the part of the Rhine delta between
Rotterdam and the North Sea. Within the Rijnmond area
of 15 3 40 km2, about one million people are located, 26.8.5 Population Characteristics
as well as a vast complex of oil, petrochemical, and The study covered the risks both to employees and to the
chemical industries and the largest harbor in the world. public. For the off-site population, a grid was used show-
A description of the industrial complex at Rijnmond has ing the number of people over an area of some 75 km2,
been given by Molle and Wever (1984). There are five which covered the whole of the Rijnmond. Populations
oil refineries with a refining capacity of some 81 Mt/year, were estimated for each 500 m2 square using data from
including the large refineries of Shell at Pernis and BP the 1971 census, updated in 1975.
Chapter | 26 Case Histories 481

TABLE 26.10 Rijnmond Report: Some Hazard Models Used


Event Model

A Emission
Flow from
A. Vessel containing liquid at atmospheric pressure
B. Vessel containing liquid above atmospheric boiling point
C. Vessel containing gas only
Two-phase flow Fauske Cude (Fauske, 1964; Cude, 1975)
B Vaporization
Spreading of liquid spill Shaw and Briscoe (1978 SRD R100)
Vaporization of cryogenic liquid
On water Shaw and Briscoe (1978 SRD R100)
On land Shaw and Briscoe (1978 SRD R100); AGA (1974)
Combined spreading and vaporization of cryogenic liquid on land
Vaporization from spill into complex bunds
Evaporation of volatile liquid Pasquill (1943), Opschoor (1978)
C Gas dispersion
Neutral density gas dispersion Pasquill Gifford
Dense gas dispersion R.A. Cox and Roe (1977)
Dispersion of jet of dense gas Ooms, Mahieu and Zelis (1974)
D Vapor cloud combustion
Initial dilution with air
Continuous release Ooms et al. (1974)
Instantaneous release Empirical relation
Ignition sources
Fire versus explosion
E Vapor cloud explosion

Vapor cloud explosion TNO (Wiekema, 1980)


F Pool fire
Flame emissivity
Liquid burning rate Burgess et al. (1961)
Flame length P.H. Thomas (1963)
Flame tilt P.H. Thomas (1965b)
Fraction of heat radiated
View factor Rein, Sliepcevich and Welker (1970)
G Jet flame
Flame length API (1969 RP 521)
Deflection effect API (1969 RP 521)
Buoyancy and liquid effects MITI (1976)
View factor Rein et al. (1970)

H Fireball
Fireball dimensions
Radius R.W. High (1968)
Duration R.W. High (1968)
View factor McGuire (1953)
482 Lees’ Process Safety Essentials

Outside the population grid, average population densi- 26.8.8 Assessed Risks
ties were used. These were required only in cases where an
exceptionally large cloud occurred. The results of the assessment include risk contours, risk
For the particular installations studied, data was to employees, risk to public (individual risk, societal risk,
obtained on the numbers of employees present during day FN tables, FN curves), and average annual fatalities.
and night and on their locations.
For toxic gas hazard, it was assumed that 1% of the 26.8.9 Remedial Measures
population is outdoors. This estimate includes an allowance
for people taking shelter. For explosion hazard, it was For each installation, recommendations were made for
assumed that 10% are outdoors. remedial measures to reduce the risks. In accordance with
the remit of the study, the proposals made were illustra-
tive rather than exhaustive and are suggestions for further
26.8.6 Mitigation of Exposure evaluation.
For the AN storage, the main measure proposed is
The principal mitigations of exposure considered were nitrogen blanketing to prevent internal explosion.
evacuation and shelter. For evacuation, the estimate of For the ammonia storage, the principal measure
the proportion evacuated is taken as a function of the suggested is the provision of a bund. Another proposal
warning time. In some cases, the proportion is estimated made is the provision of an emergency air supply at the
as 90 100%. control room to allow the operators time to close emer-
For the effect of shelter from toxic gas, the single- gency isolation valves.
exponential stage model of ventilation is used. For the chlorine storage, the main recommendations
are directed to reduction of the size of escapes from
26.8.7 Individual Assessments pipework by the use of excess flow valves.
For the LNG storage, the main measure proposed is
The individual assessments included: directed toward the prevention of escalation due to the
G Acrylonitrile storage (Paktank): The acrylonitrile brittle fracture of carbon steel from a small release of
(AN) storage is one of a number of storages at the cold liquefied gas.
Botlek terminal of Paktank, a tank storage company For the propylene storage, the installation of remote
providing bulk storage and associated services to the isolation valves is the principal measure proposed. It was
oil and chemical industries. estimated that this would reduce the total risk by half.
G Ammonia storage (UKF): The ammonia storage is Proposals are also made for modifications to the vapor
part of a complex of plants producing ammonia, nitric return system.
acid, ammonium nitrate, urea, and other fertilizers at For the hydrodesulfurizer the measures proposed are
the Pernis site of UKF. minor. It is suggested that there should be monitoring
G Chlorine storage (AKZO): The chlorine storage of hydrogen sulfide in the control room and that the
consists of five spheres associated with chlorine cell hazard of spread of DEA through the drain system be
rooms at the Botlek site of AKZO. investigated.
G LNG storage (Gasunie): The LNG storage consists of
two tanks and associated equipment at the peak shaving 26.8.10 Critiques
plant of Gasunie.
G Propylene storage (Oxirane): The propylene storage is Rijnmond Report itself contains its own critique in the
associated with the propylene oxide plant of Oxirane at form of a review by outside consultants and industrial
Seinehaven. There are two spheres, each with a capacity comment. Much of this relates to the failure and event
of 600 te. The ambient temperature was taken in the data, the hazard models, and the injury relations and
study as 15 C, that corresponds to a gauge pressure of particularly to the models.
8 bar.
G Hydrodesulfurizer (Shell): The section of the hydrode- 26.9. SAN CARLOS DE LA RAPITA
sulfurizer studied is the diethanolamine (DEA) regen-
erator. The latter is part of an absorber regenerator
DISASTER
unit which removes hydrogen sulfide from sour gas. At about 14:30 in the afternoon of July 11, 1978, a road
The hydrogen sulfide is absorbed into DEA in the tanker carrying propylene caught on fire, eventually
absorber and the fat DEA passes to the regenerator leading to a BLEVE. The incident occurred just in front
where a reversible reaction occurs releasing the hydro- of Los Alfaques camp site, The Sand Dunes, located in
gen sulfide again. the municipality of Alcanar (Tarragona, Spain) between
Chapter | 26 Case Histories 483

Barcelona and Valencia, only 3 km from the village of San The tanker vehicle was torn into four main pieces.
Carlos de la Rapita. As a result, 217 people lost their lives Some two-thirds of the tank, landed on the ground about
and approximately 600 were injured (Manich, 2008). 150 m, about 300 m from the starting point. The middle
section traveled about 100 m to the north-east, into the
camp. The cab unit traveled some 60 m along the line of
26.9.1 The Camp Site the road to the south. The front end cap fetched up about
The camp site was a triangular shaped piece of land 100 m beyond the cab.
between the coastal road and the sea, some 3 km from the
nearest township, San Carlos de la Rapita, to the north. 26.9.4 The Emergency and the Aftermath
It was 200 m in length and 10,000 m2 in area, tapering For about half an hour after the fire, there were some
from about 100 m wide at the north to 30 m at the south 200 300 people milling around the camp, many seriously
end and 60 m wide at the point of the accident. There was burned and calling for help. Private cars and taxis began
a brick wall on a concrete foundation between the road to ferry the injured to a hospital at Amposta 13 km away.
and the camp. The first ambulance came at 14v45 from the Shell oil
The legal capacity of the camp site was 260 people; drilling site at San Carlos. The municipal ambulances
however, on the day of the accident, the camp site was came at 15:05 and the fire brigade at about 15:30.
fully booked with some 800 people staying on the origi- The road was still blocked by the burning cab; victims
nal grounds as well as on the two new undeclared had to be rescued from both sides of the camp and taken
extensions. north to Barcelona or south to Valencia as circumstances
permitted. Those taken north to Barcelona received
primary medical care at points en route, while those taken
26.9.2 The Road Tanker south to Valencia, 165 km away, did not.
The road tanker was manufactured in 1973, designed with Over 100 people died outright. Others died later from
a maximum capacity of 45 m3, and a maximum allowed burns. Out of 148 cases, 122 had third degree burns on
mass at 8 bars and 4 C of 19.35 tons. It had no pressure 50% or more of the body. Either there were fewer people
relief valve and no pressure test certificate. only slightly burned or they received treatment elsewhere
At 12:05 the tanker took on a load of propylene at the (Arturson, 1981).
ENPETROL Tarragona refinery. At the moment when
the vehicle left the refinery (12:35), it was plated as 26.9.5 Lessons of San Carlos De La
suitable for a maximum load of 22 tons, and it was Rapita Disaster
actually weighed out at 23.5 tons. The driver had a
recommended route, which was the motorway, and was The most important lessons of Los Alfaques Disaster are:
provided with the toll money, but actually took the N340 1. Education on equipment, procedures, supervision, and
coast road which went past the Los Alfaques camp. training to prevent overfilling of vehicles carrying
hazardous materials.
2. Provision for pressure relief on vehicles carrying
26.9.3 The Fire and Explosion
flammable materials.
The sequence of events at the camp site is not entirely 3. Routing of vehicles carrying hazardous materials away
clear, but several eyewitness accounts are available. from populated areas and vulnerable targets.
Marshall (1987) quotes the following. A young man serv- 4. Prompt treatment of burn victims.
ing a customer in the camp shop off the main site heard
a bang. He got into his car to investigate and 2 min
after the first sound heard a severe explosion. A ‘fireball’
26.10. PIPER ALPHA
appeared on the site. A tourist in the camp restaurant At 10:00 p.m. on July 6, 1988 an explosion occurred in the
heard a ‘pop’ from a tanker on the main road and saw a gas compression module of the Piper Alpha oil production
milky cloud drifting toward him. He ran to move his car platform in the North Sea. A large pool fire took hold in
and seconds later the cloud ignited. the adjacent oil separation module, and a massive plume
It is reported that until the main ‘fireball’ event, peo- of black smoke enveloped the platform at and above the
ple stood around watching the ball of smoke from the production deck, including the accommodation. The pool
fire. When it did occur, large numbers of people, many fire extended to the deck below, where after 20 min it
scantily clad, were burned, some running into the sea to burned through a gas riser from the pipeline connection
escape or douse the flames. between the Piper and Tartan platforms. The gas from
Over 90% of the camp site was gutted by fire. the riser burned as a huge jet flame. The lifeboats were
484 Lees’ Process Safety Essentials

inaccessible due to the smoke. Some 62 men escaped, Quarters West (LQW); and the Additional Accommodation
mainly by climbing down knotted ropes or by jumping West (AAW).
from a height, but 167 died, the majority in the quarters. The control room was in a mezzanine level in the
The Piper Alpha Inquiry has been of crucial impor- upper part of D Module. It was located about one quarter
tance in the development of the offshore safety regime in of the way along the C/D firewall from the west face.
the UK sector of the North Sea. There were two flares on the south end of the plat-
form, the east and west flares, and there was a heat shield
around Module A to provide protection against the heat
26.10.1 The Company from the flares.
The main production areas were equipped with a fire
The Piper Alpha oil platform was owned by a consortium
and gas detection system. In C Module the gas detection
consisting of Occidental Petroleum (Caledonia) Ltd,
system was divided into five zones: Cl and C2 in the west
Texaco Britain Ltd, International Thomson plc, and Texas
and east halves of the module and C3, C4, and C5 at the
Petroleum Ltd and was operated by Occidental.
three compressors, respectively.
The fire water deluge system consisted of ring mains
which delivered foam to Modules A C and part of
26.10.2 The Field and the Platform Module D and at the Tartan and MCP-01 pig traps and
The Piper Alpha platform was located in the Piper field water at the condensate injection pumps. The fire pumps
some 110 miles north-east of Aberdeen. The Piper plat- were supplied from the main electrical supply, but there
form separated the fluid produced by the wells into oil, were backup diesel-driven pumps.
gas, and condensate. The oil was pumped by pipeline to
the Flotta oil terminal in the Orkneys, the condensate
26.10.3 Events Prior to the Explosion
being injected back into the oil for transport to shore. The
gas was transmitted by pipeline to the manifold compres- On July 6, there was a major work program on the platform.
sion platform MCP-01, where it joined the major gas This included the installation of a new riser for the Chanter
pipeline from the Frigg Field to St Fergus. field and work on a prover and metering loop.
There were two other platforms connected to Piper The GCM was also out of service for changeout of the
Alpha. Oil from the Claymore platform, also operated by molecular sieve driers. In consequence, the plant opera-
Occidental, was piped to join the Piper oil line at the tion had reverted to the phase 1 mode so that the gas was
‘Claymore T.’ Claymore was short of gas and was there- relatively wet.
fore connected to Piper Alpha by a gas pipeline so that it The resulting increased potential for hydrate formation
could import Piper gas. Oil from Tartan was piped to was recognized by management onshore. The increased
Claymore and then to Flotta and gas from Tartan was methanol injection rates required were calculated, the
piped to Piper and thence to MCP-01. methanol injection rates needed were some 12 times
The production deck level consisted of four modules, greater than for normal phase 2 operation.
Modules A D. Module A was the wellhead, Module B There was an interruption of the methanol supply to
the oil separation module, Module C the gas compression the most critical point, at the JT valve, between 4:00 and
module, and Module D the power generation and utilities 8:00 p.m. that evening.
module. The operating condensate injection pump was pump
Module A was about 150 ft long east to west, 50 ft B. Pump A was down for maintenance. There were three
wide north to south, and 24 ft high. The other modules maintenance jobs to be done on this pump: (1) a full
were of approximately similar size. There were firewalls 24 month preventive maintenance (PM), (2) repair of
between Modules A and B, between Modules B and C, the pump coupling, and (3) recertification of PSV 504.
and between Modules C and D (the A/B, B/C and C/D In order to carry out the 24 month PM, the pump had
firewalls, respectively); these firewalls were not designed been isolated by closing the gas operated valves (GOVs)
to resist blasts. on the suction and delivery lines but slip plates had not
The pig traps for the three gas risers from Tartan and been inserted.
to MCP-01 and Claymore were on the 68 ft level. Also With the pump in this state, with the GOVs closed but
on this level were the dive complex and the JT flash without slip plate isolation, access was given to remove
drum, the condensate suction vessel and the condensate PSV 504 for testing. It was taken off in the morning of
injection pumps. July 6. They were not able to restore the PSV that evening.
There were four accommodation modules: the East The supervisor in this team came back to the control
Replacement Quarters (ERQ), the main quarters module; room sometime before 6:00 p.m. to suspend the permit-to-
the Additional Accommodation East (AAE); the Living work (PTW).
Chapter | 26 Case Histories 485

At about 4:50 p.m. that day, just at shift changeover, Most of the personnel on the platform were in the
the maintenance status of the pump underwent a change. accommodation, the majority in the ERQ. Within the first
The maintenance superintendent decided that the 24 minute, flames also appeared on the north face of the
month PM would not be carried out and that work on the module and the module was enveloped in the smoke
pump should be restricted to the repair of the pump plume coming from the south. The escape routes from
coupling. were impassable.
About 9:50 p.m. that evening pump B tripped out At the 68 ft level, divers were working with one man
on the 68 ft level. The lead production operator and under water. They followed procedure, got the man up
the phase 1 operator attempted to restart it but without and briefly through the decompression chamber. They
success. The loss of this pump meant that with pump A were unable to reach the lifeboats, which were inaccessi-
also down, condensate would back up in the JT flash ble due to the smoke. They therefore launched life rafts
drum and within some 30 min would force a shut-down and climbed down by knotted rope to the lowest level,
of the gas plant. the 20 ft level.
The lead operator came up to the control room. It was The drill crew also followed procedure and secured
agreed to attempt to start pump A. The lead maintenance the wellhead.
hand came down to the control room to organize the elec- The oil pool in Module B began to spill over onto the
tricians to deisolate the pump. They were observed at the 68 ft level where another fire took hold. There were
pumps by the phase 2 operator and an instrument fitter, drums of rigwash stored on that level which may have
but the evidence of these witnesses was inconclusive. fed the fire.
There was no doubt that the lead operator intended to The fire water drench system did not operate. There
start pump A. was only a trickle of water from the sprinkler heads.
About 9:55 p.m., the signals for the tripping of two of The explosion disabled the main communications sys-
the centrifugal compressors in Module C came up in the tem which was centered on Piper. The other platforms
control room. This was followed by a low gas alarm were unable to communicate with Piper. They became
in zone C3 on centrifugal compressor C. Then, the third aware that there was a fire on Piper, but did not appreci-
centrifugal compressor tripped. Before the control room ate its scale. They continued for some time in production
operator could take any action a further group of alarms and pumping oil.
came up: three low gas alarms in zones C2, C4, and C5 Some 20 min after the initial explosion, the fire on the
and a high gas alarm. The operator had his hand out to 68 ft level led to the rupture of the Tartan riser on the
cancel the alarms when he was blown across the room by side outboard of the ESV. This resulted in a massive jet
the explosion. flame.
Just prior to the explosion, personnel in workshops in The emergency procedure was for personnel to report
Module D heard a loud screeching sound which lasted for to their lifeboat, but in practice, most evacuations would
about 30 s. be by helicopter and personnel would be directed from
the lifeboats to the dining area on the upper deck of the
ERQ and then to the helideck. Personnel in the ERQ
26.10.4 The Explosion, the Escalation, found the escape routes to the lifeboats blocked and
waited in the dining area. The OIM told them that a
and the Rescue Mayday signal had been sent and that he expected heli-
The initial explosion occurred at 10:00 p.m. It destroyed copters to be sent to effect the evacuation. In fact, the
most of the B/C and C/D firewalls and blew across the helideck was already inaccessible to helicopters.
room the two occupants of the control room, the control Some 33 min into the incident, the Tharos picked up
room operator, and the lead maintenance hand. the signal. ‘People majority in galley. Tharos come.
The explosion was followed almost immediately by a Gangway. Hoses. Getting bad.’
large fireball which issued from the west side of Module No escape from the ERQ to the sea was organized by
B and a large oil pool fire at the west end of that module. the senior management. However, as the quarters began
The large oil pool fire gave rise to a massive smoke to fill with smoke, people filtered out by various routes
plume which enveloped the platform from the production and tried to make their escape.
deck at the 84 ft level up. Some men climbed down knotted ropes to the sea.
The offshore installation manager made his way to the Others jumped from various levels, including the helideck
radio room and had a Mayday signal sent. at 174 ft.
The Tharos effectively took on the role of On-Scene The vessels around the platform launched their fast
Commander. Rescue helicopters and a Nimrod aircraft for rescue craft (FRCs). The first man rescued, by the FRC
aerial on-scene command were dispatched. of the Silver Pit, was the oil laboratory chemist, who, on
486 Lees’ Process Safety Essentials

experiencing the explosion, simply walked down to the installation may shelter while the emergency is brought
20 ft level. Most survivors, however, were rescued from under control and evacuation organized.
the sea. Further, it is recommended that this demonstration
The FRC of another vessel, the Sandhaven, was should be by QRA.
destroyed with only one survivor. The recommendation on the safety case includes a
At about 10:50 p.m. the MCP-01 riser ruptured, and requirement that the operator should demonstrate that it
about 11:18 p.m. the Claymore riser ruptured. The pipe has a safety management system (SMS).
deck collapsed and the ERQ tipped. By 12:15 a.m. on The report considers that the then current regulatory
July 7, the north end of the platform had disappeared. By body, the DoEn, is unsuitable as the body to be charged
the morning only Module A, the wellhead, remained with implementing the new regime and recommends the
standing. transfer of responsibility for offshore safety to the HSE.
These recommendations were accepted immediately
by the government and the new regime under the HSE
26.10.5 Some Lessons of Piper Alpha began in April 1991.
A list of some of the lessons is given in Table 26.11.
26.11. THREE MILE ISLAND
At 4:00 on March 28, 1979, a transient occurred on Reactor
26.10.6 Recommendations on the Offshore No. 2 at Three Mile Island, near Harrisburg, Pennsylvania.
Safety Regime A turbine tripped and caused a plant upset. The operators
The Piper Alpha Report makes recommendations for tried to restore conditions, but misinterpreting the instru-
fundamental changes in the offshore safety regime. ment signals, misjudged the situation, and took actions
The basis of the recommendations is that the respon- which resulted in the loss of much of the water in the
sibility for safety should lie with the operator of the reactor and the partial uncovering of the core. Radioactivity
installation and that nothing in the regime should detract escaped into the containment building. Site and general
from this. emergencies were declared.
A central feature of the regime proposed is the safety The accident at Three Mile Island (TMI) (also referred
case for the installation. In the offshore safety case, it is to initially as Harrisburg) was the most serious accident
required that the operator should demonstrate that the which had occurred in the US nuclear industry.
installation has a TSR in which the personnel on the
26.11.1 The Company
The company which operated TMI was Metropolitan
TABLE 26.11 Some Lessons of Piper Alpha Edison (Met Ed).
Regulatory control of offshore installations Safety Management
System 26.11.2 The Site and the Works
Documentation of plant TMI is situated on the Susquehanna River, some 10 miles
Fallback states in plant operation south-east of Harrisburg, Pennsylvania.
Permit-to-work systems

Isolation of plant for maintenance 26.11.3 The Process and the Plant
Training of contractors personnel The plant consisted of two pressurized water reactors
Disabling of protective equipment by explosion itself
(PWRs) of approximately 900 MWe each.
The reactor itself is a pressure vessel containing the
Offshore installations: control of pressure systems for reactor core. There are some 37,000 of these fuel pins
hydrocarbons at high pressure, limitation of inventory on
installation, emergency shut-down system, fire and explosion
arranged in clusters in the core. The core is cooled by
protection, temporary safe refuge, formal safety, use of wind the primary cooling system containing water at a high
tunnel test and explosion simulations in design pressure (,,15.2 MPa) to prevent it from boiling.
The explosion and fire phenomena: explosions in
Heat is removed from the reactor in this primary
semi-confined modules, pool fires and jet flames cooling circuit, the reactor cooling system (RCS), and used
in the steam generators to generate steam to drive a pair
Publication of reports of accident investigations
of turbogenerators. The primary cooling system consists of
two independent loops containing water at high pressure.
Chapter | 26 Case Histories 487

Since the volume of the cooling water changes with 26.11.5 The Excursion 1
temperature, it is necessary to have some free space or
‘bubble’ in the primary circuit to control the system pres- At the time of the incident, the reactor was operating under
sure. This is provided by the pressurizer, which is automatic control at 97% of its rated output. The incident
equipped with electrical heating to raise steam and thus began when water which had entered the instrument air
increase the pressure, and cold water sprays to condense line caused the isolation valves on the condensate polishing
steam and thus lower the pressure. Steam can be blown system to drift shut and the condensate booster pump to
off through a pilot-operated relief valve (PORV). The tail lose suction pressure and trip out. The main feedwater
pipe from the PORV is run to the reactor coolant drain pumps in the secondary circuit then tripped and almost
tank. immediately the main turbine tripped. The time was
An emergency core cooling system (ECCS) is pro- 4:00:37.
vided to give backup cooling. This consists of putting Valves allowing steam to be dumped to the con-
coolant into the system at a pressure greater than the denser opened, and the auxiliary feedwater pumps
normal coolant pressure. Another is the core flooding started up. The removal of heat from the RCS fell, and
system, which automatically injects cold borated water the pressure of the RCS rose. Within 3 6 s the PORV
into the system when the coolant pressure falls below set pressure of 15.5 MPa was reached and the valve
about 4.1 MPa. The third is the low pressure injection opened, but this was insufficient to relieve the pressure,
system which operates at 2.8 MPa. This takes water and at 8 s in, pressure rose to 16.6 MPa, the set point for
either from a large tank of borated water or from the reactor trip. The control rods were driven into the core
dump of the containment building and is thus capable of to stop the reaction.
keeping up recirculation indefinitely. The RCS pressure now fell below the PORV set
The reactor, the primary cooling circuit, and the steam pressure, but the valve failed to shut. However, the valve
generators are housed in a containment building which is position indicator in the control room, which actually
leak-tight. The containment is designed to retain the indicated not the position of the valve but the signal sent
contents of the primary circuit in the event of a major to it, showed the valve as closed.
leak. In the secondary circuit condensate was no longer
The turbogenerator set is housed in a separate building being returned to the steam generators. The auxiliary
which also contains the condensate and water treatment feedwater pumps were running, but the valves between the
plants. Feedwater for the secondary circuit is from the con- pumps and the steam generators were also, inadvertently,
denser. The condensate water is purified by ion exchange closed. At 1 min 45 s into the incident the steam generators
in a condensate polishing system, a package unit. boiled dry.
The water converted to steam on the secondary side in Meanwhile the RCS pressure was dropping. At 2 min
the steam generators is supplied to them by the main the pressure fell to 11 MPa at which point the first of the
feedwater pumps. These are backed up by auxiliary, or ECCS systems, the HPI system, came on and injected
emergency, feedwater pumps. Both sets of feedwater high pressure water. The liquid level in the pressurizer,
pumps are housed in the turbine building. however, was rising.
The display of liquid level was an indicator of the
level of fluid in the pressurizer, but no longer of the mass
26.11.4 Events Prior to the Excursion of water; the liquid now contained steam bubbles and had
Some 2 days prior to the incident, routine testing was thus been rendered less dense. To this extent the level
done on the valves on the auxiliary feedwater pumps. display had become misleading.
Two of the valves, the ‘twelve valves’, had inadvertently The operators became concerned that the HPI system
been left shut. was increasing the inventory of water in the RCS, and
For some 11 h before the incident, the operators had that the circuit would become full of water and thus go
been trying to clear a blockage in the condensate polishing ‘solid’. At 4 min 38 s they therefore shut down one of the
system. In order to do this, they made use of compressed HPI pumps and throttled back the other.
air. The service air line used for this purpose was cross- With water passing through the PORV, the pressure in
connected to the instrument air line. The pressure of the air the reactor coolant drain tank built up and at 7 min 45 s
in the instrument line was less than that of the water in the the reactor building sump pump came on, transferring
units and some water got into the instrument air line, water to the waste tanks in the auxiliary building.
despite the presence of a non-return valve on that line. This At 8 min the operators realized that the steam genera-
water found its way to some of the plant instruments. tors were dry. They found that the auxiliary feedwater
Another feature was a persistent slight leak from system valves were closed and opened them, thus restor-
either the regular safety valves or the PORV. ing feedwater to the steam generators.
488 Lees’ Process Safety Essentials

At 60 min the operators found that the RCS pumps particular concern. There were also a number of other
were vibrating. This was due to the presence of steam, problems, mainly associated with the large quantities of
though they did not realize this. At 1 h 14 min they shut water contaminated with low level radioactivity.
off the two pumps in Loop B to prevent damage to the On March 30, a radiation reading outside the plant, later
pumps and pipework, with possible loss of coolant, and found to be erroneous, led NRC officials to recommend
at 1 h 40 min shut off the two pumps in Loop A. evacuation. The local director of emergency preparedness
At 2 h 18 min the PORV block valve was at last shut was notified to stand by for an evacuation; he notified
off, thus stopping the loss of water from the RCS. The fire departments and had a warning broadcast that an evacu-
RCS pressure then began to rise. ation might be called. Then the NRC chairman assured the
In the 2 h since the turbine trip periodic alarms had Governor that an evacuation was not necessary. Later,
warned of low level radiation in the containment building. the Governor decided that pregnant women and pre-school
At about 2 h in, there was a marked increase in the radia- children should be evacuated. By April 1, the NRC had
tion readings. By 2 h 48 min in, high radiation levels concluded that the hydrogen bubble posed no threat but
existed in several areas. At 2 h 55 min a site emergency failed to announce this properly.
was declared.
Attempts were made to restart the RCS pumps. One
pump in Loop B operated for some 19 min but was shut
26.11.7 The Excursion 2
down again by vibration trips at 3 h 13 min. At 3 h 30 min The operators in the TMI-2 control room made a number
a general emergency was declared. of errors. Some of these were failures to make a correct
At some point, perhaps about 4 h in, the station man- diagnosis of the situation, while others were undesirable
ager appears to have realized that the reactor core had acts of intervention.
suffered damage. The first was the failure to realize that the PORV had
From 4 h 30 min attempts were made to collapse the stuck open. There were several central-indications, how-
steam bubbles in the RCS loops, but without success and ever, which pointed to the fact that the PORV was open.
at 7 h these efforts were abandoned. Closely connected with this first diagnostic failure
The operators then tried to bring in the lower pressure was the failure to recognize that the mass of primary
cooling system by reducing pressure in the RCS. They coolant water had fallen. Here the operators were misled
began at 7 h 38 min by opening the PORV block valve. by the level measurement.
At 8 h 41 min the pressure fell to 4.1 MPa at which point Another counterproductive action was the shutting
the core flooding system operated. down of the primary coolant pumps. About 60 min into
One consequence of the exposure of the core was that the incident, the pumps began vibrating.
a steam-zirconium reaction occurred, generating hydro- These failures of diagnosis and the fear of the system
gen. During the depressurization, hydrogen from the RCS going solid have frequently been characterized as illustra-
was vented into the containment building. At 9 h 50 min, tions of mindset.
there occurred in the containment building what was The operators were poorly served by the displays and
apparently a hydrogen explosion. The sprays came on for alarms in the control room.
some 6 min.
When the significance of the event in the containment
building became appreciated, concern grew about a possible
26.11.8 Some Lessons
explosion of a hydrogen bubble there. The Three Mile Island incident yields numerous lessons.
Some of these are listed in Table 26.12.
26.11.6 The Emergency and the Aftermath
The site emergency was declared at 7:00 on March 28.
26.12. CHERNOBYL
At 7:24 a general emergency was declared. On Monday, April 28, 1986, a worker at the Forsmark
The plant was contacted by the local WKBO radio nuclear power station in Sweden put his foot in a radia-
station. The radio station asked what kind of emergency tion detector for a routine check and registered a high
it was. They were told that it was a general emergency, a reading. The station staff thought they had had a radioac-
‘red-tape’ sort of thing required by the NRC when certain tive release from their plant and the alarm was raised.
conditions exist, but that there was no danger to the public. However, as reports came in of high radioactivity in
The emergency did not end with the establishment of Stockholm and Helsinki, the source of the release was
stable operation. The following day, the core damage was identified as the Soviet Union. In fact, an accident had
found to be more serious than first thought. The hydrogen occurred on Unit 4 at Chernobyl in the Ukraine some 800
bubble problem persisted over the next 8 days and caused miles from Sweden on Saturday April 26.
Chapter | 26 Case Histories 489

The reactor contained 192 te of uranium enriched to


TABLE 26.12 Some Lessons of Three Mile Island 2%. The subdivision of the uranium into a large number
Regulatory control of major hazard installations of separately cooled fuel channels greatly reduced the
risk of total core meltdown.
People-related versus equipment-related problems
The reactor was cooled by water passing through the
Formal safety assurance pressure tubes. The water was heated to boiling point and
Relative importance of large and small failures partially vaporized. The steam water mixture with a mass
steam quality of 14% passed to two separators where steam
Influence of operator in an emergency
was flashed off and sent to the turbines, while water was
Fault diagnosis by the process operator mixed with the steam condensate and fed through downco-
Accident management for major hazard installations mers to pumps which pumped it back to the reactor.
There were two separate loops each with four pumps, three
Follow-up of safety issues
operating and one standby. This system constituted the
Learning from precursor events multiple forced circulation circuit (MFCC).
Package and other ancillary units The reactor contained a stack of 2488 graphite blocks
with a total mass of 1700 te.
Display of variable of interest
An emergency core cooling system (ECCS) was pro-
Plant status display and disturbance analysis systems vided to remove residual heat from the core in the event
Abuse of instrument air of loss of coolant from the MFCC.
There was a control system which controlled the
Limitations of non-return valves
power output of the reactor by moving the control rods,
in and out.
The reactor was housed in a containment built to
withstand a pressure of 0.45 MPa. There was a ‘bubble’ con-
At 1:24 on that day an experiment to check the use of denser designed to condense steam entering the containment
the turbine during rundown as an emergency power supply from relief valves or from rupture of the MFCC.
for the reactor went catastrophically wrong. There was An important operating feature of the reactor was its
a power surge in the reactor, the coolant tubes burst and a reactivity margin, or excess reactivity. Below a certain
series of explosions rent the concrete containment. The power level the reaction would be insufficient to avoid
graphite caught fire and burnt, sending out a plume of xenon poisoning. It was therefore necessary to operate
radioactive material. Emergency measures to put out the with a certain excess reactivity. The operating instructions
fire and stop the release were not effective until May 6. stated that a certain excess reactivity was to be main-
tained; the equivalent number of control rods was 30.
Another important operating characteristic was the
26.12.1 The Site and the Works reactor’s ‘positive void coefficient’. An increase in heat
The Chernobyl nuclear power station is situated in a from the fuel elements would cause increased vaporiza-
region which at the time was relatively sparsely popu- tion of water in the fuel channels and this in turn would
lated. There were some 135,000 people within a 30 km cause increased reaction and heat output. There was
radius. Of these, 49,000 lived in Pripyat to the west of the therefore an inherent instability, a positive feedback,
plant’s 3 km safety zone and 12,500 in Chernobyl 15 km which could be controlled only by manipulation of the
to the south-east of the plant. control rods.
There were four nuclear reactors on the site.
26.12.3 Events Prior to the Release
26.12.2 The Process and the Plant
The origin of the accident was the decision to carry out a
Unit 4 at Chernobyl was designed to supply steam to two test on the reactor. Electrical power for the water pumps
turbines each with an output of 500 MWe. The reactor and other auxiliary equipment on the reactor was supplied
was therefore rated at 1000 MWe or 3200 MWt. by the grid with diesel generators as back-up. However,
Unit 4 was an RBMK-1000 reactor, a boiling water in an emergency, there would be a delay of about a min-
pressure tube, graphite moderated reactor. The reactor is ute before power became available from these generators.
cooled by boiling water, but the water does not double as The objective of the test was to determine whether during
the moderator, which is graphite. this period the turbine could be used as it ran down
The design of the reactor avoided the use of a large to provide emergency power to the reactor. A test had
pressure vessel. already been carried out without success, so modifications
490 Lees’ Process Safety Essentials

had been made to the system, and the fresh test was to own weight. Within 4 s, by 1:23:44, the reactor power
check whether these had had the desired effect. had risen, according to Soviet estimates, to 100 times the
The program for the test included provision to switch nominal value.
off the ECCS to prevent its being triggered during the At about 1:24, there were two explosions, the second
test. within some 3 s of the first.
At 1:00 on April 25 the reduction of power began. It is thought that the fuel fragmented, causing a rapid
At 13:05 Turbogenerator 7 was switched off the reactor. At rise in steam pressure as the water quenched the fuel ele-
14:00 the ECCS was disconnected. However, a request was ments, so that there was extensive failure of the pressure
received to delay shutting down since the power output tubes. The explosive release of steam lifted the reactor
was needed. It was not until 23:10 that power reduction was top shield, exposing the core. Conditions were created for
resumed. a reaction between the zirconium and steam, producing
At some stage, the trip causing reactor shut-down on hydrogen. An explosion involving hydrogen occurred in
loss of steam to Turbogenerator 8 was disarmed, apparently and ruptured the containment building, and ejecting
so that if the test did not work the first time, it could be debris and sparks were seen.
repeated. This action was not in the experimental program. The accident was aggravated by the fact that the 200
The test program specified that rundown of the turbine te loading crane fell onto the core and caused further
and provision of unit power requirements was to be bursts of the pressure tubes.
carried out at a power output of 700 1000 MWt, or 20%
of rated output. The operator switched off the local auto-
matic control, but was unable to eliminate the resultant 26.12.5 The Emergency and the Immediate
imbalance in the measurement function of the overall
automatic controls and had difficulty in controlling the
Aftermath
power output, which fell to 30 MWt. Only at 1:00 on The fire brigades from Pripyat and Chernobyl set out at
April 26 was the reactor stabilized at 200 MWt, or 6% of 1:30. The fires in the machine hall over Turbogenerator 7
rated output. Meanwhile the excess reactivity available were particularly serious, because they threatened Unit 3
had been reduced as a result of xenon poisoning. also. These therefore received priority. By 5:00 the fires
It was nevertheless decided to continue with the test. in the machine hall roof and in the reactor roof had been
At 1:03 the fourth, standby pump in one of the loops of extinguished.
the MFCC was switched on and at 1:07 the standby pump However, the heating up of the core and its exposure
in the other loop was switched on. to the air caused the graphite to burn. The residual activ-
ity of the radioactive fuel provided another source of
heat. The core therefore became very hot and the site of
26.12.4 The Release raging fire.
With the reactor operating at low power, the hydraulic During the next few days the fire raged and a radioac-
resistance of the core was less, and this combined with tive plume rose from the reactor. The accident had
the use of additional pumps resulted in a high flow of become a major disaster. It was necessary to evacuate the
water through the core. This condition was forbidden population from a 30 km radius round the plant and deal
by the operating instructions, because of the danger of with the casualties and to take a whole range of measures
cavitation and vibration. The steam pressure and water to dampen and extinguish the fire, to cover and enclose
level in the separators fell. In order to avoid triggering, the core, and to deal with the radioactivity in the sur-
the trips which would shut-down the reactor, the trips rounding area.
were disarmed. The reactivity continued to fall, and the Three evacuation zones were established: a special
operator saw from a printout of the reactivity evaluation zone, a 10 km zone and 30 km zone. A total of 135,000
program that the available excess reactivity had fallen people were evacuated.
below a level requiring immediate reactor shut-down. Accounts of the evacuation are incomplete and some-
Despite this, the test was continued. times contradictory, but what appears to have happened is
At 1:23:04 the emergency valves on Turbo generator as follows. Within a few hours of the accident, an emer-
8 were closed. The reactor continued to operate at about gency headquarters was set up in Pripyat. About 14:00 on
200 MWt, but the power soon began to rise. At 1:23:40 April 26, there was an evacuation of some 1000 people
the unit shift foreman gave the order to press the scram from within a 1.6 km zone around the plant. A much
button. The rods went down into the core, but within a larger evacuation took place about 14:00 the next day,
few seconds shocks were felt and the operator saw that when about 1000 buses were brought in and within 2 h
the rods had not gone fully in. He cut off the current to had evacuated some 40,000 people. However, it was not
the servo drives to allow the rods to fall in under their until some 9 days later that the authorities in Moscow
Chapter | 26 Case Histories 491

ordered complete evacuation from a 30 km zone around


the site. TABLE 26.13 Some Lessons of Chernobyl
A medical team was set up within 4 h of the accident, Management of, and safety culture in, major hazard
and within 24 h triage of the 100 most serious cases had installations
been affected. Adherence to safety-related instructions
A system was set up to control movements between
the zones. People crossing the zone boundaries had to Inherent safer design of plants
change clothing and vehicles were decontaminated. Sensitivity and operability of plants
Decontamination of the area outside the site was Design of plant to minimize effect of violations
complicated by the fact that the distribution of the radio-
nuclides changed with time according to the terrain, Disarming of protective systems
particularly during the first 3 4 months. Hence the full Planning and conduct of experimental work on plants
value of measures to decontaminate a particular area is Accidents involving human error and their assessment
generally obtained only temporarily.
The death toll from Chernobyl cannot be known with Emergency planning for large accidents
certainty, since most deaths are due to excess cancers. Mitigating features of accidents
Two men died dealing with the accident itself. By the end
of September the toll of dead in the USSR was 31 with
203 injured. Other European countries were also affected.
Radioactive contamination was measured across Europe.
with disastrous consequences (NOAA, Hurricane Rita—
National Climatic Data Center). Katrina is the costliest
26.12.6 The Later Aftermath hurricane in the US history.
Of those on site at the time, two died on site and a further
29 in hospital over the next few weeks. A further 17 are
permanent invalids and 57 returned to work but with 26.13.2 Hurricane Katrina
seriously affected capacity. The others on site, some 200,
were affected to varying degrees. There is apparently no On August 23, 2005, a tropical depression developed in
detailed information on the effects on the military and the southeastern Bahamas, upgrading to tropical storm
civilian personnel brought in to deal with the accident. Katrina the next day. On August 25, Katrina became a
A detailed discussion of such effects from Chernobyl category 1 hurricane and passed by the Southeastern coast
is given in the Second Watt Committee Report. For of Florida. Continuing its southwest movement, Katrina
EC countries the estimates of the NRPB in 1986 were reached its highest intensity on August 28. Katrina made
100 fatal thyroid cancers and 2000 fatal general cancers, landfall near Buras, LA on August 29 at 7:10 a.m. EDT
or 2100 fatal cancers. For the USSR the situation is much as a category 3 hurricane with winds of 125 mph. Inland,
more complex, but the report quotes an estimate by Ryin the hurricane moved near the Louisiana Mississippi bor-
et al. (1990) of 1240 fatal leukemia cases and 38,000 fatal der and through Mississippi, maintaining hurricane inten-
general cancers. sity for almost 100 miles. Katrina continue to move north
through Tennessee and on August 30, it became a tropical
depression. Katrina developed 62 tornados and produced
26.12.7 Some Lessons of Chernobyl rainfalls of up to 15 inches (Johnson, 2006).
A list of some of the lessons learned from Chernobyl, that Shortly after the hurricane hit Louisiana, the flooding
are applicable for process industry is given in Table 26.13. problems in New Orleans worsened when four levees
breached resulting in floodwaters that covered about 80%
of the city up to 20 feet high. The New Orleans flood-
26.13. HURRICANES KATRINA AND RITA waters were completely removed 43 days after Katrina
made landfall (Johnson, 2006).
26.13.1 Introduction Highways and roads were damaged or flooded, and
Despite the high hurricane activity throughout the years many businesses were destroyed causing the loss of tens of
in the Gulf of Mexico, 2005 was predominantly active in thousands of jobs. Louisiana and Mississippi were signifi-
the Gulf Coast with 28 tropical storms and 15 hurricanes, cantly affected by Hurricane Katrina; Florida, Georgia, and
of which seven were category 3 or higher (Blake et al., Alabama to a lesser degree (Knabb et al., 2006).
2007). 2005 also registered for the first time two During Katrina, the evacuation was not very effective,
category-five hurricanes, hurricanes Katrina and Rita, Federal emergency officials failed to provide ground
492 Lees’ Process Safety Essentials

transportation, and neither buses nor trains were deployed Over 1070 onshore and offshore releases were reported
(Litman, 2006). to the Incident Reporting Information System (IRIS)
Because of the approximately 1800 direct and indirect between 2005 and 2008 due to Hurricane Katrina, and
fatalities (Knabb et al., 2006), the 275,000 homes even 2 years later releases were reported due to damaged
affected, and the nearly 81 billion dollars of property or sunk structures and leaks in pipelines (Kessler et al.,
damage, hurricane Katrina is the costliest and one of the 2008; Santella et al., 2010).
deadliest hurricanes in the United States (Blake et al.,
2007; Johnson, 2006).
26.13.4.1 Offshore Industry
During Katrina and Rita, 113 platforms were destroyed,
26.13.3 Hurricane Rita 457 pipelines were damaged (BOEMRE, 2006a, 2006b),
Twenty-six days after Katrina, Hurricane Rita impacted the and 611 releases at offshore facilities were recorded between
Texas Louisiana border. Hurricane Rita developed as a 2005 and 2006 (Cruz and Krausmann, 2009). Additionally,
tropical depression on September 18, 2005. On September the damage, evacuation and loss of production affected the
20, Rita became a category-two hurricane and passed by price of oil (Cruz and Drausmann, 2008).
the Florida Keys. Rita intensified to category five the next During Katrina, more than 700 platforms and rigs were
day where a maximum wind speed of 175 mph was evacuated, 44 platforms were destroyed (mostly in shallow
reached. Rita weakened to a category three and made land- waters), and 21 were severely damaged (BOEMRE 2006a,
fall on September 24 at 3:40 a.m. EDT with winds that 2006b). The major damage occurred in the Shell platform
reached 120 mph. Inland, the winds impacted more than Mars. Four drilling rigs were destroyed, nine were severely
150 miles causing significant damage. The winds continued affected, and six were set adrift (Krausmann and Cruz,
until the Louisiana Texas Arkansas border. Rita became 2008); The most severely damage rigs were the Rowan
a tropical depression on September 25. Approximately 90 New Orleans and the Diamond’s Ocean Warwick.
tornados and rainfall of five to nine inches were reported in During hurricane Rita, 69 platforms were destroyed
Louisiana, Arkansas, Mississippi, and Alabama (NOAA, and 32 were severely damaged, mostly in shallow water.
Hurricane Rita—National Climatic Data Center; Knabb One rig was destroyed, 10 others were severely damaged,
et al., 2006). Texas and the Florida Keys were also affected and 13 were set adrift (Krausmann and Cruz, 2008);
by the floods and winds (Knabb et al., 2006). MOEMRE 2006a, 2006b. For Mobile Offshore Drilling
The damage caused by Rita is estimated at approxi- Units (MODUs), both hurricanes combined destroyed
mately $10 billion in the United States (Knabb et al., 2006; 5 drilling rigs and damaged 19 (Krausmann and Cruz,
Carpenter et al., 2006). 2008).
Of the approximately 120 deaths reported during Rita, More releases occurred on platforms during hurricane
only 7 were directly caused by the hurricane (Knabb Katrina than during Rita, probably due to the preparedness
et al., 2006; Carpenter et al., 2006). Most of the deaths for Rita or because some of the facilities from Katrina
occurred during the evacuations (Carpenter et al., 2006). remained shut-down (Cruz and Krausmann, 2009).
Due to the massive evacuation, the roadways were highly
congested preventing rapid medical responses to emer-
26.13.4.2 Onshore Industry
gencies (Carpenter et al., 2006). Additionally, fuel, food,
and water shortages as well as cars out of gas worsened Numerous onshore facilities were also affected by the
the traffic congestion. The evacuation of patients was strong winds and the surges of hurricane Katrina and
inefficient (Carpenter et al., 2006), and Hobby Airport Rita. Companies with no major damage still required
and Bush Intercontinental Airport suffered delays because approximately 2 months to return to normal operating
many employees evacuated the area (Litman, 2006). In conditions. Other facilities were severely damage and
Texas, a bus exploded killing 23 passengers, all of them remained shut down for long periods of time (Sanders,
nursing home evacuees (Carpender, 2006). Because of the 2008; Santella et al., 2010). Prior to Katrina, four refin-
extreme heat and humidity (Sanders, 2008), many people eries reduced runs and nine refineries were shutdown,
died due to heat exhaustion (Knabb et al., 2006). four of which remained closed for some time due to
the damage from the hurricane (Johnson, 2006; Santella
et al., 2010). Three natural gas facilities, the Dynegy
26.13.4 Effect on the Industry Plant of Yschoskey, LA, the Dynegy Plant of Venice,
In 2005, hurricanes Katrina and Rita destroyed 109 facilities, LA, and the BP Plant of Pascagoula, MS, were damaged
damaged 53 others, and caused a significant amount of oil by hurricane Katrina, but no releases were reported
and hazardous materials spills, both onshore and offshore (Santella et al., 2010). The most significant spill during
(Kaiser and Pulsipher, 2007; Cruz and Krausmann, 2009). Katrina occurred in the Murphy Oil refinery in the
Chapter | 26 Case Histories 493

St. Bernard Parish (Meraux) where 25,100 barrels of 26.14. BP AMERICA REFINERY EXPLOSION,
mixed crude oil were released when a partially filled stor- TEXAS CITY, TEXAS, USA
age tank was dislodged and ruptured. The spill affected
approximately 1800 homes (Santella et al., 2010). On March 23, 2005, a massive explosion occurred at the
During Rita, 15 refineries were taken offline in advance BP Texas City Refinery, in Texas City, Texas, USA. This
in Texas and Louisiana, and one of them remained shut accident resulted in 15 fatalities and 180 injuries and was
down until 1 month after the hurricane made landfall considered as one of the most catastrophic industrial
(Godoy, 2007). accident in the US history. The explosion happened at
approximately 1:20 p.m. during the startup of a hydrocar-
bon isomerization unit when a distillation tower flooded
with hydrocarbons, caused over pressurization and relief
26.13.5 Lessons Learned devices to open resulting in a geyser-like release of flam-
It is important to recognize the devastating power that mable liquid from the vent stack that was not equipped
hurricanes can have and, even more important, learn with a flare. Most victims at the time of the incident were
from others that experienced the destruction of hurricanes located in office trailers located near to the blowdown
Katrina and Rita (Blake et al., 2007; Sanders, 2008; vessel. In addition to casualties, houses and buildings
Keel et al., 2008; Krausmann and Cruz, 2008). Two key located as far as three quarters of a mile away from the
aspects should be considered to avoid the loss of lives refinery were significantly damaged. The financial losses
and property and to prevent potential releases: design to from this accident exceed $1.5 billion.
withstand hurricanes and plan in advance to minimize the
damage.
26.14.1 The Company
Texas City refinery is the company’s largest refinery
worldwide with capacity up to 10 million gallons of
26.13.6 Recommendations
gasoline per day. There are 29 oil refining units and
Key factors should be implemented to avoid loss of life 4 chemical units within its 1200 acre site with approxi-
and minimize potential damage caused by hurricanes. The mately 1800 BP workers. At the time of the incident,
following lists some practices known to provide effective BP was performing a turnaround and there were about
protection and response. 800 contractor workers onsite to support the work.
Planning: Develop hurricane preparedness plans that
consider the probability of occurrence, the damage
expected from winds and surges, and the vulnerability 26.14.2 Incident Description
of the facility. Then, identify suitable mitigations The incident at the BP refinery in Texas City, Texas,
options. Provide detailed shutdown and start-up proce- occurred on March 23, 2005. The incident occurred dur-
dures closely related to evacuation plans. Include evac- ing the restarting of the refinery’s ISOM process unit
uation plans that ensure the transportation of workers after turnaround maintenance. The ISOM unit process
and their families. Ensure the availability of hurricane consists of an Ultrafiner desulfurizer, a Penex reactor,
shelters and harden buildings, supplies, backup commu- a vapor recovery/liquid recycle unit and a raffinate split-
nication system, and storage of gasoline (Santella et al., ter. The incident started at the raffinate splitter section
2010). when the raffinate splitter tower was overfilled with
Evacuation: Have an accurate weather information liquid during the startup. Heavy raffinate was pumped
system. For offshore operations, the shutdown and evacu- from the bottom of the splitter tower and circulated
ation should start 5 7 days before the hurricane. through the reboiler furnace and back to the bottom tray.
Decrease inventory of hazardous materials. Transfer the Heavy raffinate product was taken of as a side stream and
material from storage tanks and fill the tanks with water. sent to storage. A level transmitter was installed in the
Inspect the foundation of tanks to determine the potential splitter tower and provided the reading of the liquid level
for a storage tank to be moved by the storm (Keel et al., to the control room. The tower was also equipped with
2008). alarms that indicated high liquid level.
Post-storm: Ensure that post-storm plans are followed. The raffinate splitter tower was protected by three par-
Assess the damage and report to management, govern- allel safety relief valves located in the overhead vapor
ment agencies and property insurance carriers accord- line. The relief valves were designed to open and release
ingly. Share experiences with other industries to develop primarily vapor when their set pressures were exceeded.
best practices and determine the aspects that need to be The vapor was then sent to the disposal header collection
improved (Cruz and Krausmann, 2009). system. The header collection system consisted of a
494 Lees’ Process Safety Essentials

14-inch NPS elevated pipe approximately 270 m away The splitter tower feed and bottoms pumps were shut
from the raffinate splitter tower. This section and two off and the raffinate section was filled and the circulation
other additional headers from other ISOM unit sections was shut down. The level control valve remained in the
were connected to the blowdown drum. ‘closed’ position for the following shift to resume the
Around the time of the incident, two major turn- startup.
arounds were underway in the ultracracker unit (ULC) At 6:00 a.m. in the morning shift change took place.
and the aromatics recovery unit (ARU). The Night Lead Operator had left approximately an hour
A month before the incident, the raffinate splitter before the shift change and left the Night Board Operator
section was shut down and the raffinate splitter tower was with brief explanation of the actions he had taken. During
drained, purged, and steamed-out to remove residual the shift change, the Night Board Operator provided only
hydrocarbons. At the time of the incident, almost all few details about the night shift’s raffinate section startup
scheduled maintenance work had been completed except activities to the Day Board Operator. The Day Board
for the Penex reactor which was awaiting a gasket; how- Operator read the logbook, interpreted that the liquid was
ever, BP decided to start up the raffinate splitter section. only added to the tower and was unaware that the heat
During the incident, no pre-startup safety review exchangers, the piping, and associated equipment had
(PSSR) procedure were conducted to verify the ade- also been filled during the previous shift.
quacy of the safety system and equipment in the ISOM On the morning of March 23, a series of mis-
process unit although the company policy required a communications took place with regard to whether the
formal PSSR prior to startup following a turnaround. startup could proceed or not. The Day Supervisor did not
The process safety coordinator responsible for an area of have a solid knowledge of what steps the night crew
the refinery including the ISOM unit was not familiar had completed and did not review the applicable startup
with the applicability of PSSR, and therefore, the PSSR procedure with the crew.
was not performed. Prior to sending additional liquid hydrocarbons to the
During pre-startup equipment checks, the instrumenta- raffinate section, the 8-inch NPS chain-operated valve
tion and equipment which was identified as malfunction- was open by outside operators to remove nitrogen from
ing were not repaired. During the ISOM turnaround, the splitter tower. A mis-communication occurred regard-
operations personnel informed the turnaround supervisor ing how the feed and product would be connected to the
that the splitter level transmitter and level sight glass tower. The Day Board Operator believed that the heavy
were malfunctioning. When the ISOM unit was operating, raffinate product should not be sent to storage; however,
this equipment could not be repaired because the block the outside operators believed that they were instructed
valves needed to isolate the level transmitter were not to send the light raffinate product to storage.
leaking. The isolation valves were replaced during the At 9:40 a.m. prior to the restart of the splitter tower,
turnaround but the BP supervisors decided not to repair the Day Board Operator opened the level control valve to
the level transmitter due to insufficient time to com- 70% output and approximately 12,000 barrel per day
plete the job in the existing turnaround schedule. The (bpd) of heavy raffinate flowed out of the tower for about
supervisors planned to repair the level transmitter after 3 min. He closed the level control valve to 0% output, but
the startup. the flow indication showed that the flow dropped only to
In the days prior to the raffinate splitter startup, one of 4300 bpd rather than 0 bpd although no heavy raffinate
the operations personnel found that a pressure control flowed out of the tower. Then the Day Board Operator
valve was malfunctioning both during the unit shutdown resumed the startup, restarted the circulation and intro-
and the equipment checks. In addition, the functionality duced the feed on the splitter tower without considering
check of all alarms and instruments required prior to the written procedure which indicated the exact stage of
startup were not completed. the startup.
The raffinate section in the ISOM unit was started The Day Board Operator believed that the verbal
during the night shift on March 22, 2005; then it was instruction required his action to maintain higher level in
stopped to be re-started during the next shift. The Night the tower to protect downstream equipment although the
Lead Operator did not record the completed steps for the startup procedure called for level control valve to be set
filling process of the raffinate section, and therefore, the in ‘automatic’ and 50% to establish heavy raffinate flow
operators on the next shift did not have any records on to the storage. The Day Board Operator noticed the level
what had been done. transmitter indicated 97% and thought it was normal.
In the early morning on March 23, 2005, the raffinate At around 10:00 a.m., as much as 20,000 bpd of raffinate
splitter tower was filled beyond the set points of both alarms feed was pumped into the tower and the output flow
to a level reading of 99% on the transmitter; however, only showed to be 4100 bpd although the Day Board Operator
one alarm was activated and sounded at 3:09 a.m. was aware that the level control valve was shut. CSB
Chapter | 26 Case Histories 495

investigation post-incident concluded that there was likely opened and liquid raffinate flowed to the raffinate splitter
no flow leaving the tower at that period. disposal header collection system.
As the tower was filled, the tower level transmitter The Day Board Operator and the outside operators
showed the liquid level less than 100%. The level sight noticed that the splitter tower pressure had spiked dramat-
glass was reported nonfunctional for several years. ically so the day board operator tried to troubleshoot the
When the unit was being heated, the Day Supervisor problem by reducing the fuel gas to feed preheater and
who had ISOM-experience left the plant due to a family fully opened the level control valve to raffinate storage.
emergency. The second Day Supervisor had very little However, data showed that the incoming feed to the
experience with the ISOM unit and therefore did not get raffinate splitter tower had not been stopped. The data
involved in the ISOM startup. At 9:55 a.m. two burners also indicated that the three safety relief valves remained
which pre-heated the feed flowing into the heater were lit fully open for about 6 min. The amount of material and
heating the liquid at the bottom of the tower. About an pressure in the tower overhead reduced due to the heavy
hour and a half later, two additional burners were lit. raffinate rundown and raffinate splitter reflux flow and
Post-incident analysis by the CSB determined that the thus the pressure dropped inside the tower overhead and
level in the tower was actually 20 ft when the level read- the safety relief valves closed. Approximately 196,000
ing was 93% and when the level decreased to 88% the liters of flammable liquid flowed from the overhead
level in the tower was actually 30 m. vapor line through the safety relief valves to the collec-
The pressure inside the tower rose to 33 psig tion header and then discharged to the blowdown drum
(228 kPa) because of the compression of the nitrogen by during the opening of the safety relief valves. Because the
the liquid in the raffinate section. As a result of this high blowdown drum completely filled, the flammable liquid
pressure, the outside operators opened the 8-inch NPS flowed into the ISOM unit sewer system.
chain-operated valve that vented directly to the blowdown The geyser-like flammable liquid was discharged from
drum in order to reduce the pressure in the tower. During the stack and fell to the ground. The first notification that
the startup, the temperature of the reboiler return to the the blowdown was overflowing was sent via radio. In
tower was as high as 307 F (153 C), the temperature response to this message, the Board Operator and Lead
increased at a rate of 73 F (23 C) per hour during the Operator used the computerized control system to shut
period 10 a.m. to 1 p.m. The heat to furnace was then the flow of the fuel to the heater. Other operators left the
reduced by decreasing the fuel gas flow to the furnace. satellite room and ran toward an adjacent road to redirect
At this time, the level transmitter at the raffinate splitter the traffic away from the blowdown drum. These actions
showed 80% but the tower level was actually 43 m. At were required by the company emergency response proce-
12:42 p.m., the splitter level control was opened to 15% dure. The ISOM operators did not have sufficient time to
output. The valve was opened five times but the heavy sound the emergency alarm before the explosion. When
raffinate flow had only begun at 12:59 p.m. The heavy the explosion occurred, hundreds of alarms registered in
raffinate flowed out the tower at 20,500 bpd at 1:02 p.m the control room sounded but the operations personnel
and 2 min later it spiked to 27,500 bpd. The level of did not have sufficient time to assess the situation.
liquid at that time was 48 m but the level transmitter read- A number of witnesses reported seeing geyser-like liquid
ing continued to decrease up to 78%. The total liquid in flowing down the 6.1-m stack.
the 52-m tower was decreased due to the opening of the The liquid hydrocarbon was released and a flammable
heavy raffinate control valve; however, the level of the vapor cloud was formed and reached an estimated area of
liquid at the top of the column kept increasing. 200,000 ft2 (18,581 m2). The vapor cloud covered this
Heating from the furnace caused contact between the area in a very short interval due to the direct dispersion
bubbles of hot vapor rising through the column and the from evaporation prior to ignition as well as the burning
cold liquid. The raffinate leaving the bottom of the column by the subsonic flames through the flammable cloud. The
circulated onto the heat exchanger, heated the feed and impact of the falling liquid through the stack onto process
resulted in rapid heating of the column section above the equipment, structural components, and piping enhanced
feed inlet. The increase in the overall column temperature the liquid fragmentation into relatively small droplets and
gave rise to the decrease of the liquid hydrocarbon density promoted the evaporation and formation of the vapor
and therefore significantly increased the level of the liquid cloud.
inside the tower. The entire column was eventually heated The wind also contributed in enhancing the mixing
approaching the boiling point of the liquid so that vapor between the flammable vapors with air. The wind at the
was accumulated and no longer condensed. time of the incident was reported to be at 5 miles (8 km)
The hydrostatic head in the line increased due to the per hour in to the southeast direction.
entrance of the liquid to the overhead line and exceeded Several potential ignition sources were identified in
the set pressures of the safety relief valves. The valves this incident, but the idling diesel pickup truck was
496 Lees’ Process Safety Essentials

suspected to be the most likely ignition point. Once While there were no serious injuries or fatalities
ignited, the flame rapidly spread through the flammable resulting from this incident, 43 people were directly
vapor cloud pushing the gas ahead of it to create a blast affected and about 2000 people were evacuated from their
pressure wave. The flame was accelerated by the combi- houses (Mannan, 2009; BMIIB, 2006).
nation of congestion/confinement and the flammable mix The Buncefield incident resulted in significantly dam-
so that intensified the blast pressure in several areas. aging many commercial as well as residential properties
close to its vicinity. The Buncefield Depot was close to
the Maylands Industrial Estate having around 630 firms
26.14.3 Aftermath and Emergency with 16,500 people. The estimated losses in the Maylands
businesses ranged between $207 270 million.
The explosion killed 15 contract employees who worked in Significant environmental damages were reported as an
or near the trailer site between the ISOM and NDU unit. aftermath of the incident in terms of the water pollution
Sixty-six workers were seriously injured and 114 others and the decreased air quality in the affected areas to very
required medical attention. Of those who were seriously high cost of clean-up and environmental remediation.
injured, 14 were BP employees and the rest were contrac-
tor employees from 13 different organizations.
The blast caused severe damage within the ISOM unit
and the surrounding parking areas. About 70 vehicles and 26.15.2 Causes of the Incident
40 trailers were damaged; 13 trailers were completely Lack of learning culture: Following the incident, it was
destroyed. Buildings surrounding the ISOM unit were also claimed that cold gasoline had never resulted in a vapor
impacted by the blast and were characterized by broken cloud explosion previously. However, a review of litera-
windows, cracked wall, damaged door, bent metal structure ture and past incidents (Kletz, 1986) indicate otherwise
and dispersal of interior contents. that there are many incidents with significant similarities
The actions of the emergency response teams were with the Buncefield incident. According to Kletz (2001c),
fast and effective in helping the victims. Member similar types of incidents keep occurring. Hence, he
companies of Texas City Industrial Aid System (IMAS) concludes that organizations have no memory and there is
assisted with fire hose line and search-and-rescue of the a greater need for digging into the past by performing
victims. incident investigations, root cause analysis of near-misses,
and estimating probabilities of major incident occurrences
(Mannan, 2011).
26.15. BUNCEFIELD Inadequate and non-functional equipment: The fuel
level in the tank was monitored by using the servo gauge
26.15.1 Description of the Incident technique, which is based on level and temperature mea-
Buncefield site was collectively operated by a number of surements. Servo gauges are utilized because they require
different companies. However, the incident was initiated in less maintenance and it is possible to also monitor the tank
one of the tanks of Hertfordshire Oil Storage Ltd. which level manually. Additionally, the temperature of the fuel
was a COMAH site housing about 200,000 metric tons of was measured using a temperature sensor (Powell, 2006).
cold gasoline inventory (BMIIB initial report, 2006). Human factors: In this incident, actively assessing the
On December 10, 2005 around 7 p.m. in the evening, process hazards and analyzing the risks were not priori-
one of the tanks entitled Tank 912 began receiving about tized (DNV report, 2008). Similarly, the possibility of the
296,644 ft3 unleaded gasoline fuel. The tank gauge records formation of vapor clouds because of overfilling was
at this point indicated that there was 38,105 ft3 of unleaded never investigated and modeling of the resulting vapor
motor fuel. The initial flow rate of the tank was 19,423 ft3/ pressure, heat radiation, and domino effects were also not
h. Around 3:00 a.m., the Automated Tank Gauging System considered (DNV report, 2006).
showed that the level of the tank was two-thirds full. By Siting of occupied vehicles and running engines:
5:35 a.m. on the same day, the fuel in Tank 912 began to Mannan (2009) suggests that based on the accident inves-
overflow with an increase in the temperature. Within the tigation, the average height of the vapor cloud would
following 20 min, the flow rate to Tank 912 had also have been at least 6.6 ft. The footage received from
increased to 34,430 ft3/h. Around 6 a.m. about 300 metric CCTV and the description of eye witnesses corroborate
tons of unleaded gasoline had overflowed from Tank 912 that the vapor cloud could be given a conservative
resulting in a vapor cloud explosion. The initial explosion estimate of about 1,610,000 ft2 which would indicate that
was followed by several other explosions which eventually the cloud volume could range between 5,650,000 ft3 and
engulfed 23 tanks in the tank farm causing a very large 10,590,000 ft3 (Mannan, 2009; Powell, 2006; Gant and
scale fire. Atkinson Report, 2006).
Chapter | 26 Case Histories 497

26.15.3 Lessons Learned from the Incident of 8 year of preparation following President John F.
Kennedy’s mission statement which ignited the space
Lessons from past incidents: In case of the Buncefield race with the Soviet Union. Apollo 11 used a lunar orbit
explosion, there was an underlying belief that cold gaso- rendezvous (LOR) to achieve the lunar landing. This
line and its vapors could not explode in the open air. required a rocket to achieve orbit with the moon and a
Unfortunately, this was a belief shared by many oil and separate lunar module which would make the trip to the
gas companies owning tank farms. Authorities and com- surface and return the astronauts to the orbiting space-
panies alike were unaware of other similar incidents such craft. Early in the space program, most engineers at the
as Newark, New Jersey in 1983, Naples, Italy in 1995, National Aeronautics and Space Administration (NASA)
and St. Herblain, France in 1991. initially preferred a direct ascent mode of travel. This
Conducting proper hazard and risk assessments: The would utilize one craft to both reach the moon and land
study of the Buncefield incident has indicated that gener- on its surface. Subsequently, the same craft could return
ally insufficient attention to incident precursors and leading the astronauts to Earth. This was set aside in favor of
indicators result in devastating explosions (DNV, 2009; LOR, but would be revived later as the Space Shuttle
Mannan et al., 2010). Emphasis is lacking in properly Program.
understanding different processes and their hazards. There
is also a lack of awareness of the properties of chemical
compounds utilized in process operations. 26.16.2 Columbia’s Final Flight
Vapor cloud formation and explosion mechanism: In the Columbia began its 113th (STS-107) flight on January 16,
case of the Buncefield incident, the BMIIB Third Progress 2003. There were seven crew members aboard. The mis-
Report discusses the spread of the vapor cloud and its depth sion consisted of 16 days in orbit around Earth for scien-
of 3.3 ft (1 m) in the area between bund A and the loading tific research in areas including microgravity and various
gantry, and 16.4 23 ft down Three Cherry Trees Lane. Earth science disciplines.
Eye witnesses also reported the presence of a gasoline mist During the shuttle’s ascent, a piece of thermal insula-
and odor along with automatic revving of cars despite their tion foam separated from the external fuel tank and struck
ignitions being turned off. Powell (2006) in the third the left wing. Shuttle fuel tanks contain liquid hydrogen at
BMIIB reports that smaller explosions followed one major 2423 F (2 253 C) (also oxygen) for propulsion and are
explosion, but only the major explosion was recorded fitted with insulation to prevent ice formation and melting
seismically. The delay in subsequent explosions after the of the external tanks. Previous shuttle launches had also
initial explosion could indicate that they were because of an shown foam separation, an event that came to be termed
internal tank explosion and fuel leakage from the pipework ‘foam shedding’. In these previous missions, the results
and tank damage. had been primarily benign, although the resulting damage
It is safe to assume that generally, tank farms have to Atlantis during mission STS-27 was of enough concern
plenty of ignition sources in their surroundings. Ignition that the mission’s commander, Robert Gibson, expected
sources might result due to improper procedures in hot the shuttle to be destroyed upon reentry.
works such as welding and cutting during maintenance Columbia began its scheduled reentry on February 1,
operations. Sometimes, ignition sources are also present 2003 with the expectation that the damage sustained dur-
because of smoking near tanks. ing takeoff would not affect the landing. As the shuttle
High overpressures generation in open air settings: entered the atmosphere, excessive heating and strain were
The preliminary assessment of the blast damage by the observed on the left wing’s leading edge. Heating and
investigation team suggested the generation of large over- hydraulic sensors began to fail and TPS tiles and other
pressures. They found that the overpressure values were of debris were shed as the shuttle began to break apart over
the range of 10.2 14.5 psi in the car parks of the buildings the southwest United States. Just before 9:00 a.m.,
in the facility. The car park location most likely was the Mission Control lost contact with the crew. Witnesses to
area where the gasoline vapors would have ignited. the crash notified NASA and a recovery team was sent
to find the crew and shuttle remains (CAIB, 2003).
26.16. SPACE SHUTTLE COLUMBIA
DISASTER 26.16.3 Accident Analysis
26.16.1 Development of the Space Foam shedding became an accepted phenomenon of shuttle
launches by NASA. It was observed on at least 6 prior
Shuttle Program missions and yielded no significant accidents, despite the
On July 20, 1969, Apollo 11 made the first lunar landing concerns raised during STS-27. Normalization of deviance
by a manned space craft. The landing was the culmination is a term applied to a safety environment in which events
498 Lees’ Process Safety Essentials

initially considered outside the accepted safety standards as organizational causes as the foundation for what went
occur without serious incident. The events are then viewed wrong during these missions. Adams (2004) refers to the
as acceptable and the bounds of safety are expanded to ‘destructive organizational dynamics’ as root causes that
include this new risk. This type of mentality was applied led to the inevitable accident.
to NASA after both the Challenger and Columbia disasters. The organizational structure of NASA at the time of
After Columbia entered orbit, NASA engineers began the accident was basically a dependent structure that left
issuing requests to the Department of Defense (DOD) for room for potential conflicts of interest. The overlapping
imaging to assess the extent of the wing damage as well jurisdiction of the safety programs at different levels by
as information obtained by the astronauts’ own inspec- the Division Chief blurred the lines of responsibility and
tions. These requests were not met and NASA manage- authority, and instead, produced faulty channels of com-
ment chose to focus on reentry scenarios rather than a munications, lack of corporate memory, flawed manage-
complete analysis and potential repair of the damage. ment and hierarchy, and a generally poor safety culture
In actuality, the foam had significantly damaged the (CAIB, 2003).
TPS, leaving the left wing susceptible to the extreme heat One of the major recommendations by the Rogers
it experiences entering the Earth’s atmosphere (.2800 F). commission was the creation of a Headquarters Office of
Hot atmospheric gases were able to penetrate the leading Safety, Reliability, and Quality Assurance, which would
edge of the wing and damage the internal structure leading fulfill the most significant gaps in the safety culture of
to a loss of hydraulics and failure of the wing’s tempera- the organization. This shift in activities was believed to
ture sensors. This led directly to the breakup of the shuttle involve an impartial party that would re-establish effec-
as it shed various components across Texas and Louisiana. tive communication channels and would have authority
over safety, reliability, and quality issues involved in
shuttle missions. However, NASA neglected full imple-
26.16.4 Other Factors Considered mentation of this change from the beginning as evidenced
The CAIB’s report included the immediate foam strike as when independent power was not provided as specified.
the cause of the accident, a detailed analysis of the safety This created a false sense of safety that compromised the
environment at NASA and the decision-making process perception of management having a fully operational pro-
leading up to and during STS-107. The safety culture at gram with a low probability of failure, rather than a pro-
NASA was cited as a significant root cause. gram in the development stages (CAIB, 2003).
Budgetary problems were also noted. US competition After the Columbia disaster, the commission recom-
with the Soviet Union for space supremacy largely ended mended the establishment of an independent Technical
with the Cold War, making access to funding and support Engineering Authority to ensure a shift in the organiza-
for NASA more difficult. tion’s pathway to destruction. The objective of this
The overall risk management strategy of NASA was measure was to guarantee the unbiased supervision of
criticized. Reliance on simulation software not pro- the division by removing their dependency on program
grammed for this specific situation was determined to be resources by providing guidance on technical standards,
an ineffective method to assess risk. hazard analysis, and risk assessment (CAIB, 2003).
Many recommendations were generated as a result of
the accident. Some of the most important include: 26.16.6 Implications for the Future of
G Proper camera placement upon shuttle launch. Human Space Flight
G Reduction of schedule pressures focusing on available
The CAIB noted several faults with actions taken by
resources.
NASA officials prior to and during the Columbia
G Creation of two independent entities, one for technical
accident. However, it acknowledged that the shuttle is
engineering and one for safety. not inherently unsafe in its current form and sufficient
G Development of the ability to analyze and repair shuttle
changes to the safety practices could maintain the
wings in space. In Columbia’s case, the crew had
program as safe for flight. Potential causes for future
enough oxygen and supplies to last for 30 days, giving
disasters should be constantly sought out while never
another shuttle time to rendezvous for a repair attempt.
settling for a culture that begins to accept anomalies as
acceptable risk.
26.16.5 The Accident’s Organizational
Causes 26.16.7 Recommendations
The investigations that followed both the Challenger and The CAIB report (CAIB, 2003) laid out 29 specific
Columbia disasters identified physical (technical) as well recommendations that addressed many areas, from shuttle
Chapter | 26 Case Histories 499

operation to safety culture at NASA. In relation to shuttle 26.17.2 The Site and the Works
operation, they specify the development and implementation
of an inspection plan to be conducted before the shuttle’s The Macondo well was located in Mississippi Canyon
return flight and also call for the ability to image the shuttle’s Block 252, about 48 miles off the coast of Louisiana. BP
exterior in orbit to determine its ability for safe return. jointly acquired the 10 year lease to this area in the cen-
Specific safety environment recommendations were tral Gulf of Mexico on March 19, 2008 from the Minerals
also included. Shuttle flight schedules should be realistic Management Service (MMS). BP owned 65% of the
and appropriate with available resources and the ability to lease, while Anadarko owned 25%, and MOEX Offshore
alter the schedule as the need arises. owned 10%. BP was the lease operator.
The Macondo well was an exploration well, as there
was promise of oil to be found below the surface;
however, BP had limited information about the geology, as
26.17. DEEPWATER HORIZON it was its first well on Block 252. The well was designed
At approximately 9:49 p.m. on April 20, 2010, a cloud to be completed into a production well in the case that
of flammable hydrocarbons released from a US Gulf of sufficient hydrocarbons were found to justify production.
Mexico deepwater well encountered an unknown source The exploration plan for the lease was accepted by
of ignition and caused two explosions in rapid succession MMS on April 6, 2009 and approved in revised form
and a subsequent fire. The events that even now continue 10 days later. The drilling permit was approved by the
to unfold uncover many underlying problems in the MMS on May 22, 2009. The semi-submersible drilling
process safety considerations for all involved parties. unit Marianas was used to begin drilling the well in
The incident occurred on the Deepwater Horizon October of 2009. In November of 2009, Hurricane Ida
mobile offshore drilling unit owned and operated by caused extensive damage to Marianas and caused it to go
Transocean located about 48 miles off the coast of offline.
Louisiana. The lease was held by BP who contracted Deepwater Horizon was used as the substitute to the
Transocean to drill the well for their Macondo Prospect. damaged Marianas after the MMS approved a revised
Eleven workers were killed in the explosion and a further drilling application for the well on January 14, 2010.
17 were injured. The fire that was caused by the explo- Deepwater Horizon had been a familiar drilling unit to
sions burned for 36 h before the rig sank on April 22, BP, who had leased it from Transocean for 9 years at the
2010. The oil spill that resulted from the uncontrolled beginning of the Macondo drilling project. It had been
wellbore flow incident developed into a matter of world- used for approximately 30 drilling projects by BP, the
wide interest as it became the largest oil spill to originate most recent before Macondo being the Kodiak appraisal
in US waters in history. well, also off the coast of Louisiana in Mississippi Block
771 (SubseaIQ).1

26.17.1 The Companies 26.17.3 Events Prior to the Explosions


The companies generally considered to be directly BP initially estimated the cost of completion of the
involved in the operations of the well are BP, the lessee Macondo well to be $96.2 million and the expected time
of the drilling area, and Transocean, the owner of of completion was 51 days. At the time of the incident,
Deepwater Horizon and drilling contractor. Related par- BP was $58 million over budget and 6 weeks behind
ties responsible for various parts of the drilling operations schedule. This situation likely contributed to decisions
include Halliburton, who was the cementer for the job, made up to the day of the blowout that served to save
and Cameron, the manufacturer of the blowout preventer. time and money perhaps in lieu of safety considerations.
Numerous other contractors in diverse fields had employ-
ees aboard at the time of the accident, including compa-
nies supporting other drilling operations, administrative 26.17.4 Events Prior to April 19
support, and housekeeping (USCG).
The drilling operations were designed by BP, but the Events that led to or foreshadowed the blowout began at
physical work of actually running the well was contracted the time that Deepwater Horizon arrived at the site to
to Transocean, as is common practice (Presidential replace the damaged Marianas unit.
Commission). These employees worked under the direc-
tion of two ‘Well Site Leaders’ provided by BP for guid- 1 Information in sections 26.17.6 and 26.17.7 is primarily taken from the
various reports for the Deepwater Horizon incident; specifically, BP
ance and direction, as well as for communication with (2010), Presidential Commission (2011), DNV (2011), and Transocean
onshore BP engineers. (2011).
500 Lees’ Process Safety Essentials

Yellow pod of BOP leaks: Problems with the blowout that end. However, in the end, the process of cementing
preventer (BOP) itself began within the first month of comes down to the art of running a blind operation that
operation when on February 23 a pilot valve leak of depends on the prior science and engineering to succeed.
1 gpm was noticed on the yellow pod of the blowout In cementing, the crew is not able to see the results of
preventer. Operation was switched to the blue pod, and the job, so it must rely upon well-designed procedures,
leaking was reduced. and testing to ascertain whether the job has been done
Original drill pipe sheared: A well control event properly.
occurred on March 8 that caused the drill pipe to become Preparing for temporary abandonment: The final plan
stuck in the formation and subsequently abandoned. for temporary abandonment was not delivered to the rig
Lost circulation event: On April 9, at 18,198 ft below until the morning of the procedure, after it had already
sea level, much shallower than the objective depth of begun. It included an abnormally large amount (425 bar-
20,200 ft, the pressure exerted by the drilling mud began rels, or twice the normal amount) of spacer to be used to
to damage the fragile formation. complete the displacement, which was formed from two
Selection of long string casing: With the lost circulation highly viscous water-based pills so that the pill material
event, it was obvious that the original design objectives could be discharged overboard legally instead of taken to
for the casing had to be reevaluated to ensure that the shore for disposal. It differed from the previous plans
formation would be able to withstand future production in the ordering of events and the presence of a negative-
operations. pressure test. As noted by BP, it did not include a formal
Selection of adequate number of centralizers: The risk assessment of the annulus cement barriers, which
original design of the long-string casing was determined did not conform to their own guidelines and may have
to require 16 or more centralizers, while only 6 were identified problems with the cement job.
available from the supplier at the time. Further calcula- Pressure tests: Both positive- and negative-pressure
tions and simulations after the decision between long- tests were carried out as a part of the procedure for tem-
string and liner showed that the well needed more than 6 porary abandonment. The positive-pressure test is meant
centralizers and could even need as many as 21. to examine the mechanical integrity of the casing under
pressure, while the negative-pressure test is designed to
ensure that the bottom-hole cement job can keep the
26.17.5 Events on April 19 and 20 hydrocarbons from flowing out of the well.
The events beginning with the final casing run, completed Between the positive pressure test and the negative
on April 19, and continuing through the cementing pressure test, the sequence for temporary abandonment
process, pressure tests, and preparations for temporary was followed and the drill pipe was run into the ground
abandonment had a direct effect on the incident itself, and the mud was displaced with seawater to a point
as problems began to compound. above the BOP. At about 5 p.m. the negative-pressure
Final casing run: The final casing run actually started test began.
early on April 18, and was not completed until mid- The crew opened the bleed valve to initialize the pres-
afternoon on April 19. The number of stabilizers used sure to 0 psi, measured on the drill pipe. The procedure is
was the aforementioned 6 as opposed to the recom- to bleed the line to 0 psi, close the valve, and check that
mended 21. it stays at zero. However, for the first trial the pressure
Float collar: The final casing run was completed to a normalized at between 240 and 266 psi and, furthermore,
terminal depth of 18,304 ft and the shoe track was fitted when the crew closed the valve, it shot up to between
with the float collar installed at the top and a reamer shoe 1250 and 1262 psi. The annular preventer was tightened
at the bottom. After the casing run was installed, nine to stop a suspected leak that was causing the fluid level
attempts were made to establish circulation. The final in the riser to drop. When the line was bled again, 0 psi
attempt at establishing circulation was deemed successful was reached; however, when the well was shut in, once
at a peak pressure of 3142 psi; however, this pressure was again the pressure spiked, this time to 773 psi. One more
far greater than the expected pressure from the manufac- attempt was made, and once again the pressure fell to
turer’s guidelines and BP’s procedures, which specified zero and spiked when the valve was closed.
500 700 psi at a fluid circulation rate of 5 8 barrels per The results of these tests may have been explained
minute, which was also not attained. incorrectly (though Transocean disputes that this was ever
The cement job: Cementing is figuratively regarded as used as an explanation) by attributing the pressure to the
a combination of science and engineering with an element ‘bladder effect’, where mud in the riser exerts pressure on
of art. The engineering tells the engineer what properties the annular preventer, which transfers to the drill pipe,
are needed for the cement to complete the objective, thereby giving incorrect readings. The crew decided to
while the science allows one to design the cement toward carry out another negative-pressure test, but this time on
Chapter | 26 Case Histories 501

the kill-line where, they reasoned, there would be no 6 min, but in the time it took to test the fluid, the pressure
bladder effect. in the drill pipe rose by 246 psi, indicating flow from the
The test was carried out on the kill line. After bleed- underbalanced well. This is because the seawater initially
ing the pressure down to 0 psi, the crew closed the valve surrounding the drill pipe was being pushed up from
and waited for 30 min without observing any flow or below by the remaining drilling mud below the pipe,
pressure change, though the pressure on the drill pipe was which, in turn, was being pushed up by the flowing
still 1400 psi. The negative-pressure test was deemed a hydrocarbons from the pay zone.
success even though the difference in the pressures was It is not known why the pressure was not monitored
never explained. The validity of using the kill line as the during the sheen test. At any rate, the pumps were
pressure monitor point is a point of dispute in the reports, restarted at 9:14 p.m., obscuring the pressure data from
with Transocean saying that it was not valid and BP say- the well.
ing it was valid, but the kill line may have been plugged Displacement continues: The displacement was con-
or a valve may have been left closed. tinued after the sheen test was deemed to be acceptable.
Mud displacement: When the negative-pressure test Flow was diverted overboard to discard the spacer, and
was accepted, the next step was to displace the rest of the four pumps were turned on to displace the rest of the
drilling mud with seawater and then add the cement plug to fluid. However, one of the pumps was connected to the
block off the well temporarily. The flow of hydrocarbons kill line and this happened before the kill line valve was
into the wellbore was first restricted by the hydrostatic open; at 9:18 p.m., pressure in the line rose to over
pressure of the dense drilling mud on the formation, 7000 psi and the emergency relief valve opened.
but would be less restricted by the less dense seawater. Three of the four pumps were taken off-line in order to
Theoretically, if the negative pressure test is correct, there determine which had failed and to subsequently deal with
should be no flow from the formation into the well at this the repair. When the pump in need of repair had been
point; if there is flow, it is called a kick and it indicates that ascertained, it began to be repaired while the other two
there is a serious problem with the integrity of the well. came back on-line. This caused a normal and expected
When displacement activities commenced at about increase in pressure, but within the next 5 min an unex-
8 p.m., all signs looked normal. The pressure in the pected increase in kill line pressure occurred.
drill pipe was decreasing, as expected, and the mud was Before the blowout: The increase in pressure was
being redirected to different holding tanks as necessary to not noticed until about 9:30 p.m., when it had reached
accommodate the volume coming in from the wellbore. 833 psi. The pumps were shut down to investigate this
However, according to the Presidential Commission anomaly, and the pressure fell accordingly. However,
report, as the operation went on, mud was also directed after the initial fall, the pressure began to build back up.
from the sand traps and trip tank (though not simulta- Between 9:31 and 9:34, the pressure on the drill pipe
neously) to the holding tanks used for the displaced mud. increased by approximately 560 psi and began to stabilize.
Because the holding tanks were being filled from several At this point, in an attempt to bleed the excess pressure,
locations at the same time, the measurement of the vol- a floorhand opened a drill pipe valve. The pressure was
ume of displaced mud to check for a kick would have relieved partially, but only momentarily before it crept up
been much more complicated or even impossible. again and stabilized, though at a slightly lower pressure
The well became underbalanced at about 8:52 p.m., than initially. At this point, the well was producing
meaning that the pressure exerted by the seawater and between 60 and 70 barrels per minute.
mud could not balance the pressure of the fluids in the At roughly 9:38, the pressure in the drill pipe began to
formation, causing flow of hydrocarbons into the well- drop at a constant rate, indicating that the hydrocarbons
bore. It was estimated that by 9:08, 39 barrels of fluid were pushing through the annulus past the drilling mud
from the formation had found its way into the mud being toward the top of the riser.
displaced, but there was no indication that this was recog- Influx of hydrocarbons: After the pressure drop, the
nized by the well crew. This has widely been attributed to influx of hydrocarbon was just a matter of a short delay.
the distraction of preparing the next operation, the surface Between 9:40 and 9:43, the drilling mud displaced by the
plug. But before the addition of the surface plug, a sheen hydrocarbons spewed onto the drilling floor. At 9:41 the
test was performed. annular preventer of the BOP was activated and, though it
Sheen test: At 9:08 p.m., the spacer fluid between the closed, failed to fully seal the well. The crew concurrently
drilling mud and the seawater displacing it was consid- decided to reroute the flow to the mud-gas separator to pre-
ered to have come to the top of the riser, and pumping vent it from spilling on the drilling floor. Another option,
was shut down to conduct a sheen test, or an inspection rerouting the fluid overboard, was not chosen.
of the fluid to detect any free oil in the spacer before Unfortunately, the mud-gas separator (MGS) was not
discharging it to sea. The test itself only took about designed for such high flow rates, and was quickly
502 Lees’ Process Safety Essentials

overwhelmed. The fluid being diverted soon began to


pour from its outlet lines. One minute after this event, at TABLE 26.14 Some Lessons from the Deepwater
about 9:47, natural gas from the well began to be vented Horizon Incident
from the MGS through goosenecked vents that angled Relative priority of safety and production
flow back toward the surface of the platform. Gas alarms
Management of change in drilling procedures
sounded, but were not equipped to trip any equipment
that could be a source of ignition. Also at this time, the Review process
crew activated the variable bore ram mechanism of the Lost time versus higher risk
BOP, which temporarily sealed the annulus.
Frequent changes
This evasive action was too late. At 9:49, main power
generation engines started to overspeed. The rig lost Management splitting and hierarchy in operations with more
power. About 5 s later flammable gas ignited and the first than one company
explosion occurred; only 10 s later another explosion Adequacy of and adherence to company procedures and
rocked the platform. 11 workers lost their lives in direct guidelines
conjunction with the explosions and subsequent fire. Planning for an emergency
Training
26.17.6 Lessons from the Deepwater Lessons learned from similar near-misses
Horizon Incident Procedures for maintenance
Summarized below are a series of Lessons Learned based Update of engineering drawings
on the four incident investigation reports. The incident
Design of blowout preventer and automatic blowout
reports, in general, found very similar lessons to be
prevention system
learned from Deepwater Horizon. Management and pro-
cedural problems were commonly cited as underlying Use of simulation data and other data-at-hand
causes, but there were also many circumstantial causes Interpretation of test results
of the accident to learn from as well. No accident report
Identification of hazardous situations in a timely fashion
attributes the incident to a single root cause. Some causes
from which lessons can be learned developed over Design of offshore gas detection systems
extended amounts of time, such as the failure to update Decision-making under stress
engineering drawings and maintenance failures, while
other causes developed during a short duration leading up
to incident—for example, the decisions made based upon
test data (Table 26.14).
action and the MMS was rebranded as BOEMRE. This
26.17.7 Impact reorganization would also lead to the establishment of the
Bureau of Ocean Energy Management, for the manage-
The Deepwater Horizon blowout, explosions and fire, ment of leasing and permitting for energy resources, and
sinking, and oil spill gathered worldwide attention, not
the Bureau of Safety and Environmental Enforcement,
only for the disaster itself, but for the sheer size and dura- which would be in charge of safety activities such as
tion of the incident. It was the largest accidental marine inspections. The US government also put a temporary ban
oil spill in the US history with impacts on the environ- on deepwater drilling, and issued new requirements for
ment, economy, and regulations. Short term environmen- all offshore drillwells. As part of this the voluntary API
tal impacts are easily inferred from the many reports and SEMP program, is now required for offshore operations.
images of shores and animals covered in oil from the gulf Yet to be determined are the long-term impacts of
coast, while various initiatives are assessing the long-term Deepwater Horizon on the regulatory atmosphere of the
impact. President Obama insisted that BP set up a $20
gulf and the exact impact that the disaster will have on
billion compensation fund, which was managed by a gov- the parties in question.
ernment official, to compensate those economically
impacted by the spill.
One of the early impacts on regulation was the reorga- ACRONYMS
nization of the Minerals Management Service. In the ACMH Advisory Committee on Major Hazards
aftermath of the disaster, this reorganization was put into
Chapter | 26 Case Histories 503

Amoco American Oil Company VCE vapor cloud explosion


API American Petroleum Institute VGS vent gas scrubber
ATSDR Agency for Toxic Substances and Disease Registry
BG British Gas REFERENCES
BLEVE boiling liquid expanding vapor explosion
BOEMRE Bureau of Ocean Energy Management, Regulation Assheton R., 1930. History of Explosions. Charles L. Storey Company
and Enforcement for Institute of Makers of Explosives, Wilmington, DL.
CAAA Clean Air Act Amendments Assini, J., 1974. Choosing welding fittings and flanges. Chem. Eng. 81
CCPS Center for Chemical Process Safety (September), 90
CIMAH Control of Industrial Major Accident Hazards ten Berge, W.F., 1985. The toxicity of methylisocyanate for rats.
Regulations 1984 J. Hazard. Mater. 12 (3), 309.
CISHC Chemical Industry Safety and Health Council Bhushan, B., Subramanian, A., 1985. Bhopal: what really happened?
CSB Chemical Safety and Hazard Investigation Board Business India February 25 March 10, 102.
DMV diaphragm motor valve Blake, E.S., Rappaport, E.N., Landsea, C.W., 2007. The Deadliest,
DNV Det Norske Veritas Costliest and Most Intense United States Tropical Cyclones from
EC European Community 1851 2006 (and Other Frequently Requested Hurricane Facts).
EIDAS Explosion Incidents Data Service NOAA, Technical Memorandum NWS-TPC-5, 43 pp.
EPA Environmental Protection Agency Bolton, L., 1978. What happened in Seveso. Chem. Eng. 85
HSE Health and Safety Executive (September), 104C.
HSEES Hazardous Substances Emergency Events Braun, R., Frilling, M., Schönbucher, A., 1999. Simulation of a reaction net-
Surveillance work in a semibatch reactor. Chem. Eng. Technol. 22 (11), 919 923.
ICFTU International Confederation of Free Trade Unions Browning, B., Searson, A.H., 1989. The lessons of the Thessaloniki oil
IChemE Institution of Chemical Engineers terminal fire. Loss Prevention and Safety Promotion 6, 39.1.
ICMESA Industrie Chimiche Meda Societa Azionara Buncefield major incident investigation board explosion mechanism
LNG liquefied natural gas advisory group report, 2006.
LPG liquefied petroleum gas Buncefield MIIB—Final Report, 2006. ,http://www.buncefieldinvestiga-
MARS Major Accident Reporting System tion.gov.uk/reports/index.htm#final/.. Access date: October 3, 2013.
MCA Manufacturing Chemists Association Burgess, D.S., Strasser, A., Grumer, J., 1961. Diffusive burning of liquid
MCC methylcarbamoyl chloride fuels in open trays. Fire Res. Abs. Rev. 3, 177.
MHIDAS Major Hazard Incidents Data Service Carpenter, S, Bennett, E, Peterson, G., 2006. Scenarios for ecosystem
MIC methyl isocyanate services: an overview. Ecol. Soc.29.
MMA Methyl Methacrylate Carson, P.A., Mumford, C.J., 1979. Analysis of incidents involving major
MOC Management of change hazards in the chemical industry. J. Hazards Mater. 3, 149 165.
MRS MIC refining still (Bhopal) Carson, P.A., Mumford, C.J., 2002. Hazardous Chemical Handbook. sec-
MSS MIC storage system ond ed. Elsevier, London.
NIHHS Notification of Installations Handling Hazardous Cattabeni, F., Cavallaro, A., Galli, G. (Eds.), 1978. Dioxin, Toxicological
Substances Regulations 1982 and Chemical Aspects. Halsted Press, New York, NY.
NTSB National Transportation Safety Board Comer, P.J., 1977. The dispersion of large-scale accidental releases such
NTSIP National Toxic Substance Incidents Program as Seveso. Royal Meteorological Society Meeting on Atmospheric
OECD Organization for Economic Cooperation and Surface Exchanges of Pollution, London.
Development Cox, R. A., Roe, D. R., 1977. A model of the dispersion of dense vapour
OSD Offshore Safety Division clouds. Loss Prevention and Safety Promotion 2, 359.
OSHA Occupational Safety and Health Administration Cruz, A.M., Krausmann, E, 2009. Hazardous-materials releases from
PHA process hazard analysis offshore oil and gas facilities and emergency response following
PVC polyvinyl chloride hurricanes Katrina and Rita. J. Loss Prev. Process Ind. 22,
PVH process vent header 59 65.
RIDDOR Reporting of Injuries, Diseases and Dangerous Cude, A.L., 1975. The generation, spread and decay of flammable vapour
Occurrences Regulations 1985 clouds. In: Course on Process Safety Theory and Practice,
RVVH relief valve vent header Department of Chemical Engineering, Teesside Polytechnic,
SRD Safety and Reliability Directorate Middlesbrough.
SRV safety relief valve Cullen, J., 1985. What can we learn from Pemex? Chem. Eng. (London).
TCB 1,2,4,5-tetrachlorobenzene 418, 21.
TCDD TCDD 2,3,7,8-tetrachlorodibenzo-p-dioxin Davenport, J.A., 1987. Gas plant and fuel handling facilities: an insurer’s
TCP 2,4,5-trichlorophenol view. Plant/Operations Prog. 6, 199.
TNO Toegepast-Natuurwetenschappelijk Onderzoek Davis, L.N., 1979. Frozen Fire. Friends of the Earth, San Francisco, CA.
TNT trinitrotoluene Doyle, W.H., 1969. Industrial explosions and insurance. Loss Prev. 3, 11.
UCC Union Carbide Corporation Doyle, W.H., 1972a. Instrument connected losses in the CPI. Instrum.
UCIL Union Carbide India Ltd Technol. 19 (10), 38.
504 Lees’ Process Safety Essentials

Doyle, W.H., 1972b. Some major instrument-connected CPI losses. Kessler, R.C., Galea, S., Gruber, M.J., Sampson, N.A., Ursano, R.J.,
Instrumentation in the Chemical and Process Industries 8. Instrument Wessely, S., 2008. Trends in mental illness and suicidality after
Society of America, Pittsburgh, PA. Hurricane Katrina. Mol. Psychiatry. 13, 374 384. Available from:
Dyfed County Fire Brigade, 1983. Report of the Investigation into the http://dx.doi.org/10.1038/sj.mp.4002119.
Fire at Amoco Refinery August 30, Carmarthen, Dyfed. Kier, B., Muller, G., 1983. Handbuch Störfalle. Erich Schmidt, Berlin.
Eisenberg, N.A., Lynch, C.J., Breeding, R.J., 1975. Vulnerability Model: Kimbrough, R., Krouskas, C.A., Carson, M.L., Long, T.F., Bevan, C.,
A Simulation System for Assessing Damage Resulting from Marine Tardiff, R.G., 2010. Human uptake of persistent chemicals from con-
Spills. Rep. CG-D-136 75. Enviro Control Inc., Rockville, MD. taminated soil: PCDD/Fs and PCBs. Regul. Toxicol. Pharmacol. 57,
Fauske, H.K., 1964. The discharge of saturated water through tubes. 43 54.
Seventh National Heat Transfer Conference. American Institute of Kimmerle, G., Eben, A., 1964. On the toxicity of methyl isocyanate and
Chemical Engineers, New York, p. 210. its quantitative determination in air. Arch. Toxikol. 29, 235.
Gant, S., Atkinson, G., 2006. Buncefield Investigation: Dispersion of the Kletz, T., 2003. Still Going Wrong!: Case Histories of Process Plant
Vapour Cloud Health and Safety Laboratory Report, CM/06/13. Disasters and How They Could Have Been Avoided. Butterworth-
Glasstone, S., 1964. The Effects of Nuclear Weapons. Atomic Energy Heinemann publications, Burlington MA 01803, USA.
Commission, Washington, DC. Kletz, T., 2009. What Went Wrong: Case Histories of Process Plant
Gugan, K., 1979. Unconfined Vapour Cloud Explosions. Institution of Disasters and How They Could Have Been Avoided. fifth ed
Chemical Engineers, Rugby. Butterworth-Heinemann publications, Burlington MA 01803, USA.
Haastrup, P., Brockhoff, L., 1990. Severity of accidents with hazardous Kletz, T.A., 1986. Accident reports and missing recommendations. Loss
materials. A comparison between transportation and fixed installa- Prev. Safety Promot. 5, 201.
tions. J. Loss Prev. Process Ind. 3 (4), 395. Kletz, T.A., 1988. Learning from Accidents in Industry. Butterworths,
Haastrup, P., Brockhoff, L.H., 1991. Reliability of accident case histories London.
concerning hazardous chemicals. An analysis of uncertainty and Kletz, T.A., 1990. Critical Aspects of Safety and Loss Prevention.
quality aspects. J. Hazard. Mater. 27, 339. Butterworth, London.
Hagon, D.O., 1986. Pemex (letter). Chem. Eng. (London). 421, 3. Kletz, T.A., 1993. Lessons from Disaster: How Organizations Have No
Ham, B., 2005, A review of spontaneous combustion incidents. In: Aziz, N. Memory and Accidents Recur. Gulf Professional Publishing, UK.
(Ed.), Coal 2005: Coal Operators’ Conference, University of Kletz, T.A., 2001a. An Engineer’s View of Human Error. third ed.
Wollongong & the Australasian Institute of Mining and Metallurgy, pp. Institution of Chemical Engineers, Rugby.
237 242. Kletz, T.A., 2001b. Learning from Accidents. third ed. Butterworth-
Harvey, B.H., 1976. First Report of the Advisory Committee on Major Heinemann, Oxford, UK and Woburn, MA.
Hazards. HM Stationery Office, London. Kletz, T.A. 2001c. Some problems and opportunities that have been
Harvey, B.H., 1979. Second Report of the Advisory Committee on overlooked. Proceedings of the Mary Kay O’Connor Process Saf.
Major Hazards. HM Stationery Office, London. Cent. Annu. Symp. College Station,Texas.
Harvey, B.H., 1984. Third Report of the Advisory Committee on Major Knabb, R.D., Rhome, J.R., Brown, D.P., 2006. Tropical cyclone report,
Hazards. HM Stationery Office, London. Hurricane Katrina, 23 30, August 2005. National Hurricane
Hay, A., 1976a. Seveso: the aftermath. Nature. 263, 537. Center, 43 pp. Available online at http://www.nhc.noaa.gov/pdf/
Hay, A., 1976b. Seveso: dioxin damage. Nature. 266, 7. TCRAL122005_Katrina.pdf.
Hay, A., 1976c. Seveso: seven months on. Nature. 265, 490. Krausmann, E., Cruz, A.M., 2008. Natech disaster: when natural
Hay, A., 1976d. Seveso: solicitude. Nature. 266, 384. hazards trigger technological accidents. Spec. Issue Nat. Hazards. 46
Hay, A., 1976e. Toxic cloud over Seveso. Nature. 262 (August), 636. (2).
Hay, A., 1981. Seveso: the intrigue and the infighting. Nature. 290 Lees, F.P., 1976. The reliability of instrumentation. Chem. Ind. March, 195.
(March), 71. Lees, F.P., 1980. Loss Prevention in the Process Industries, vol. 12.
Hay, A., 1982. The Chemical Scythe. Lessons of 2,4,5-T and Dioxin. Butterworths, London.
Plenum Press, New York, NY. Lenoir, E.M., Davenport, J.A., 1993. A survey of vapor cloud explo-
Health and Safety Executive (HSE), 1978. Canvey: An Investigation of sions: second update. Process Saf. Prog. 12 (1), 12-33.
Potential Hazards from Operations in the Canvey Island/Thurrock Lewis, D.J., 1980. Unconfined vapour cloud explosions: historical per-
Area. HM Stationery OfficeHSE), 1978, London. spective and predictive method based on incident record. Prog.
Health and Safety Executive (HSE), 1981. Canvey: A Second Report. Energy Combust. Sci. 6, 151.
A Review of the Potential Hazards from Operations in the Canvey Litman, T., 2006. Lessons from Katrina and Rita: what major disasters
Island/Thurrock Area Three Years After Publication of the Canvey can teach transportation planners. J. Trans. Eng. 132, 11 18.
Report. HM Stationery OfficeHSE), 1981, London. Available online at: http://scitation.aip.org/teo; http://www.vtpi.org/
Health and Safety Executive (HSE), 1989. The Fires and Explosion at katrina.pdf.
BP Oil (Grangemouth) Refinery Ltd. HM Stationery OfficeHSE), Mcguire, J.H., 1953. Heat Transfer by Radiation. Fire Research Special
1989, London. Report 2. HM Stationery Office, London.
High, R.W., 1968. The Saturn fireball. Ann. N.Y. Acad. Sci. 152 (Art. Mahoney, D.G., 1990. Large Property Damage Losses in the
1), 441. Hydrocarbon-Chemical Industries: A Thirty-Year Review. thirteenth
Howard, W.B., 1985. Seveso: cause-prevention. Plant/Operations Prog. ed. M&M Protection Consultants, New York, NY.
4, 103. Manich, O., 2008. Disenö de las bases de una logı́stica aplicada a desas-
Kaiser, M.J., Pulsipher, A.G., 2007. Generalized functional models tres y catástrofes en el ámbito de la Provincia de Barcelona. Anexo
for drilling cost estimation. SPE Drill Compl. 22 (2), 67 73, D. Department de Projected d’Enginyeria. Universitat Politècnica de
SPE-98401-PA. Catalunya. Spain.
Chapter | 26 Case Histories 505

Mannan, M.S., et al., 2009. Analysis of the Buncefield oil depot explo- Rice, A.P., 1982. Seveso accident: dioxin. In: Hazardous Materials
sion: explosion modeling and process safety perspective. Proceedings Spills Handbook, G. F. Bennett, F. S. Feates, and I. Wilder, eds.,
of the Mary Kay O’Connor 2009 International Symposium. McGraw-Hill, New York, 11 44.
Mannan, M.S., 2011. The Buncefield explosion and fire lessons learned. Richardson, J., Ham, B.W., 1996. Incidence of spontaneous combustion
Proc. Safety Prog. 30, 138 142. Available from: http://dx.doi.org/ in Queensland underground coal mines. Department of Mines and
10.1002/prs.10444. Energy Internal Report.
Mansot, J., 1989. Incendie du Depot Shell du Port Edouard Herriot a’ Lyon Richardson, T., 1991. Learn from the Phillips explosion. Hydrocarbon
les 2 et 3 Juin 1987. Loss Prevention and Safety Promotion 6, 41.1. Process. 70 (3), 83.
Margerison, T., Wallace, M., Hallenstein, D., 1980. The Superpoison. Rijnmond Public Authority, 1982. Risk Analysis of Six Potentially
Macmillan, London. Hazardous Industrial Objects in the Rijnmond Area: A Pilot Study.
Marshall, V.C., 1980. Seveso, an analysis of the official report. Chem. Springer.
Eng. (London). 358, 499. Riley, R.V., 1979. Accidents with pipelines in the USA: selected
Marshall, V.C., 1987. Major Chemical Hazards. Ellis Horwood, case histories and a review of the activities of the National
Chichester. Transportation Safety Board. Third International Conference
Nash, J.R., 1976. Darkest Hours. Nelson Hall, Chicago, IL. on Internal and External Protection of Pipes. British
Occupational Safety and Health Administration (OSHA), 1990. Phillips Hydromechanics Research Association, Cranfield,
66 Company Houston Chemical Complex Explosion and Fire. Bedfordshire.
OHSAOSHA), 1990, Washington, DC. Sambeth, J., 1983. What really happened at Seveso. Chem. Eng. 90
Ooms, G., Mahieu, A.P., Zelis, F., 1974. The plume path of vent gases (May), 44.
heavier than air. Loss Prevention and Safety Promotion 1, 211. Sanders, R.E., 2008. Hurricane Rita: an unwelcome visitor to PPG
Opschoor, G., 1978. Rep. 78-0834. TNO, Apeldoorn. industries in Lake Charles, Louisiana. J. Hazard. Mater. 159 (1),
Orsini, B., 1977. Parliamentary Commission of Inquiry on the Escape of 58 60.
Toxic Substances on July 10, 1976 at the ICMESA Establishment Santella, N., Steinberg, L.J., Sengul, H., 2010. Petroleum and hazardous
and the Consequent Potential Dangers to Health and the material releases from industrial facilities associated with hurricane
Environment due to Industrial Activity. Final Report, Rome. Katrina. Risk Analyis. 30 (4), 635 649, Published Online: March
Orsini, B., 1980. Seveso (Translation of Official Italian Report by 16, 2010.
Health and Safety Executive). Health and Safety Executive, London. Scott, J.N., 1992. Succeeding at emergency response. Chem. Eng. Prog.
Parker, R.J., 1975. The Flixborough Disaster. Report of the Court of 88 (12), 62.
Inquiry. HM Stationery Office, London. Skandia International, 1985. BLEVE! The Tragedy of San Juanico. First
Pasquill, F., 1943. Evaporation from a plane free-liquid surface into a 2 (November).
turbulent air stream. Proc. R. Soc. Ser. A 182, 75. Shaw, P. and Briscoe, F. (1978) Evaporation from spills of hazardous
Pesatori, A.C., Baccarelli, A., Consonni, D., Lania, A., Beck-Peccoz, P., liquids. UKAEA, SRD R100, Risley, UK, May 1978, p333. In
Bertazzi, P.A., et al., 2008. Aryl hydrocarbon receptor-interacting Marshall, V.C. (1987) Major Chemical Hazards. Ellis Horwood Ltd,
protein and pituitary adenomas: a population-based study on subjects ISBN 085312969X.
exposed to dioxin after the Seveso, Italy, accident. Eur. J. Endocrinol. Slater, D.H., 1978. Vapour clouds. Chem. Ind. 6 (May), 295.
159, 699 703. Theofanous, T.G., 1981. A physicochemical mechanism for the ignition
Pesatori, A.C., Consonni, D., Bachetti, S., Zocchetti, C., Bonzini, M., of the Seveso accident. Nature. 291 (June), 640.
Baccarelli, A., et al., 2003. Short- and long-term morbidity and Theofanous, T.G., 1983. The physicochemical origins of the Seveso
mortality in the population exposed to dioxin after the Seveso accident. Chem. Eng. Sci. 38 (1615), 1631.
accident. Industrial Health. 41, 127 138. Thomas, P.H., 1963. The size of flames from natural fires. Combustion
Pietersen, C.M., 1985. Analysis of the LPG Incident in San Juan 9, 844.
Ixhuatepec, Mexico City, November 19, 1984. Rep. 85-0222. TNO, Thomas, P.H., 1965. Fire Spread in Wooden Cribs, Part III: the Effect
Apeldoorn, the Netherlands. of Wind. Fire Res. Note 600. Fire Research Station,
Pietersen, C.M., 1986a. Analysis of the LPG disaster in Mexico City, Borehamwood.
November 19, 1984. Gastech. 85, 112. Union Carbide Corporation, 1985. Bhopal methyl isocyanate incident.
Pietersen, C.M., 1986b. Analysis of the LPG disaster in Mexico City. Investigation Team Report. Danbury, CT.
Loss Prevention and Safety Promotion. 5, 21 31. University Engineers Inc., 1974. An Experimental Study on the
Pietersen, C.M., 1988. Analysis of the LPG-disaster in Mexico City. Mitigation of Flammable Vapor Dispersion and Fire
J. Hazard. Mater. 20, 85. Hazards Immediately Following LNG Spills on Land. AGA
Powell, T. 2006. The Buncefield Investigation: Third Progress Report. Proj. IS-100-1
Report prepared for the Buncefield Major Incident Investigation by Vervalin, C.H. (Ed.), 1964. Fire Protection Manual for Hydrocarbon
the Health and Safety Executive (HSE) and the Environmental Processing Plants. Gulf, Houston, TX.
Agency (EA). Vervalin, C.H., 1973. Fire Protection Manual for Hydrocarbon
Redmond, T.C., 1990. Piper Alpha AU: 176 the cost of the lessons. Processing Plants. second ed. Gulf, Houston, TX.
Piper Alpha: Lessons for Life Cycle Management. Institution of Wilkinson, N.B., 1966. Explosions in History. The Hagley Museum,
Chemical Engineers, Ruby, 113 p. Wilmington, DE.
Rein, R.G., Sliepcevich, C.M., Welker, J.R., 1970. Radiation view fac- Wiekema, B.J., 1980. Vapour cloud explosion model. J. Haz. Materials
tors for tilted cylinders. J. Fire Flammability 1, 140. 3 (3), 221.

You might also like