United States Office of Pollution Prevention EPA 744-R-94-002
Environmental Protection and Toxic Substance Febiuary1994
Agency (7406)
EPA Flammable Gases And
Liquids And Their Hazards
. X _' Recycled/Recyclable
' ; V \ Pnnteo on paper that contains
•<" I
-------
FLAMMABLE GASES AND LIQUIDS AND THEIR HAZARDS
Office of Pollution Prevention and Toxics
U.S. Environmental Protection Agency
Washington, D.C. 20460
February 1994
-------
ACKNOWLEDGEMENTS
The preparation of this report has been made possible through the efforts of the Office of Pollution
Prevention and Toxic Substances "Physical and Chemical Property Hazards" Workgroup over the many years
of its existence. The many contributors addressed the challenging task of reviewing, analyzing and summarizing
the myriad existing models and empirical data on flammable chemicals and their hazards. A major
consideration in this work was the need to express the hazards of flammable chemicals as correctly and simply
as possible for use by risk managers. Finally, the Environmental Protection Agency Science Advisory Board
provided comments that were helpful in preparation of the final draft of this document.
Environmental Protection Agency Report Workgroup*
and Contributors
*Roger Garrett, Workgroup Co-Chair
*Paul Tobin, Workgroup Co-Chair
* Russell Farris, Workgroup Co-Chair
*Paul Anastas, Workgroup Co-Chair
*Robert Andrei
Paul Bickart
Kathy Bishop
*Gail Froiman
* Robert Lenahan
*Craig Matthiessen
*Fred Metz
Jerry Newsome
*Nhan Nguyen
*Ward Penberthy
Paul Quillen
* Vanessa Rodriguez
Research Contributors
Pamela Bridgen, ICF Incorporated
Maravene Edelstein, ICF Incorporated
David Goldbloom-Helzner, ICF Incorporated
Deborah Shaver, ICF Incorporated
-------
1
TABLE OF CONTENTS
Page
Executive Summary ix
1.0 Introduction 1
2.0 Background 1
3.0 Existing Classification Systems, Regulations, and Codes 3
3.1 DOT Classifications .* 3
3.2 NFPA Classification Systems 3
3.3 EEC Classification Scheme 4
3.4 Regulations and Codes Applicable to Flammable Chemicals 4
4.0 Past Fire and Explosion Incidents Involving Flammable Chemicals 5
4.1 Acute Hazardous Events Data Base 5
4.2 Accidental Release Information Program 6
4.3 OSHA Data Base 6
4.4 Fatal Hazardous Materials Accidents Database 7
4.5 Major Hazard Incident Data Service (MHIDAS) 7
4.6 M & M Protection Consultants Data Base 8
5.0 Accident Consequence Modeling 8
5.1 WHAZAN Methodology 9
5.2 ARCHIE Methodology 9
5.3 The "Yellow Book" Methodology 10
5.4 AIChE-Sponsored Course Materials 10
5.5 The "Green Book" Methodology 10
5.6 PHAST Methodology 11
5.7 Multi-Energy Models 11
5.8 New Jersey Method for Fireballs 12
6.0 Types of Flammable Chemical Fires and Explosions 12
6.1 Vapor Cloud Explosions 12
-------
11
TABLE OF CONTENTS
6.1.1 Effect of Quantity on Probability of Vapor
Cloud Explosion 14
6.1.2 Confinement and Mixing in Vapor Cloud Explosions 15
6.1.3 Energy of Vapor Cloud Explosions 15
6.1.4 Type of Ignition 15
6.1.5 Delayed Ignition Effects 15
6.2 Vapor Cloud Fires 17
6.3 BLEVEs and Resulting Fireballs : 17
6.4 Pool Fires 20
6.5 Jet Fires 20
6.6 Projectiles 20
7.0 Accident Factors and Model Inputs for Consequence Analysis 25
7.1 Model Inputs 25
7.1.1 Chemicals and Quantities Chosen for Modeling 25
7.1.2 Instantaneous and Prolonged Releases 26
7.1.3 Initial Conditions Prior to Release 26
7.1.4 Meteorology 26
7.1.5 Surface Roughness 27
8.0 Hazard Criteria 27
8.1 Hazard Criteria for Vapor Cloud Explosions 27
8.2 Hazard Criteria for Vapor Cloud Fires 28
8.3 Hazard Criteria for BLEVEs, Pool Fires, and Jet Fires 28
9.0 Modeling Results 29
9.1 Vapor Cloud Explosion Results 29
9.1.1 Modeling Results for Four Hydrocarbons 29
9.1.2 Modeling Using Other Meteorological Conditions 33
9.1.3 Modeling Prolonged Releases 33
9.2 Comparison of Vapor Cloud Explosion Data and WHAZAN Results 33
9.3 Vapor Cloud Fire Results 36
9.4 BLEVE Results 42
. 9.5 Pool Fire Results 45
9.6 Jet Fire Results 52
-------
m
TABLE OF CONTENTS
10.0 Findings 52
References : 58
Appendix A. Additional Information on Classification Systems for Flammable Chemicals A-l
A.1 Definitions Related to Classification of Flammable Substances A-l
. »
A.2 DOT Classifications A-2
A.3 NFPA Classifications A-2
A.4 EEC Indicative Criteria A-3
A.5 OSHA Regulations for Service Stations A-3
Appendix B. Methodologies for Modeling B-l
B.I WHAZAN B-l
B.I.I General Description B-l
B.1.2 WHAZAN Pressurized Release Outflow B-l
B.1.3 WHAZAN Two-Phase Outflow B-2
B.1.4 WHAZAN Gas Outflow B-2
B.1.5 WHAZAN Adiabatic Expansion Outflow B-3
B.1.6 WHAZAN Evaporation from Liquid Pool B-4'
B.1.7 WHAZAN Pool Fires B-4
B.1.8 WHAZAN Jet Fires B-6
B.I.9 WHAZAN Vapor Cloud Explosions (based on Buoyant Plume
Dispersion) B-7
B.1.10 WHAZAN Fireballs/BLEVEs B-8
B.2 ARCHIE B-9
B.2.1 General Description B-9
B.2.2 ARCHIE Pressurized Liquid Release B-9
B.2.3 ARCHIE Estimating Pool Size B-10
B.2.4 ARCHIE Evaporation of Liquid Pool B-ll
B.2.5 ARCHIE Gas Discharge from Pressurized Vessel B-12
B.2.6 ARCHIE Vapor Cloud Fire Model B-12
B.2.7 ARCHIE Unconfined Vapor Cloud Explosion B-14
B.2.8 ARCHIE Tank Overpressurization Explosion Model B-14
B.2.9 ARCHIE Fireball Model B-15
B.2.10 ARCHIE Liquid Pool Fire Model B-16
B.3 Yellow Book B-20
B.3.1 General Description B-20
B.3.2 Vapor Cloud Explosion Calculation for Gases B-20
-------
IV
TABLE OF CONTENTS
B.3.2.1 Quantity of Vapor Contributing to Explosion B-21
B.3.2.2 Determination of Damage Circles B-21
B.3.3 Yellow Book Vapor Cloud Fire Calculations for Gas B-22
B.3.4 Vapor Cloud Fire Calculations for Gases B-24
B.3.5 Pool Fire Calculations for Gases and Liquids B-26
B.3.5.1 Atmospheric Coefficient of Transmission B-26
B.3.5.2 The Geometric View Factor F * B-26
B.3.5.3 Intensity of Radiation of the Fire B-28
B.3.6 Vapor Cloud Fire for Liquids B-28
B.4 AIChE-Sponsored Course Materials B-30
B.4.1 General Description B-30
B.4.2 Vapor Cloud Fire Calculation for Gases B-30
B.4.3 Vapor Cloud Fire Calculations for Liquids B-31
B.4.4 Pool Fire Calculations for Liquid and Gases B-31
B.5 Green Book B-34
Appendix C. Hazard Criteria C-l
C.I Explosion Overpressure C-l
C.I.I Damage and Injury from Blast Waves C-l
C.1.2 Distance Determination for Different Overpressures C-9
C.2 Flammability Limits C-9
C.3 Heat Radiation C-12
C.3.1 Heat Radiation Levels Causing Death and Injury C-12
C.3.2 Distance Determination for Different Heat Radiation Levels C-15
C.3.2.1 BLEVEs C-15
C.3.2.2 Pool Fires C-15
C.4 References C-18
Appendix D. Comparison of Results of Modeling by Different Methods and
Results of Varying Release Scenarios and Meteorological Conditions D-l
D.I Vapor Cloud Explosion Results D-l
D.2 Vapor Cloud Fire Results D-2
D.3 BLEVE Results D-4
-------
V
TABLE OF CONTENTS
Page
D.4 Pool Fire Results D-4
D.5 Jet Fire Results .' D-5
D.6 Instantaneous Release Compared to Prolonged Release D-5
D.7 Moderate Meteorology Compared to Worst Case Meteorology D-5
»
Appendix E. Inputs for Modeling E-l
Appendix F. Additional Information from Accident Databases F-l
F.I Additional Information from the Acute Hazardous Events Data Base F-3
F.2 Additional Information from the ARIP, OSHA, M & M Accident Databases F-6
Appendix G. Contacts with Experts G-l
-------
vi
LIST OF EXHIBITS
Exhibit Page
1 Chemicals Known to Have Been Involved in Vapor Cloud Explosions 13
2 Sample Quantities Involved in Vapor Cloud Explosions 13
3 Time Before Ignition 16
4 Distance to Ignition 17
5 Incidents Involving BLEVEs in the M & M Database 19
. *
6 Projectiles from Explosion Incidents 21
7 Sample Calculations for Projectiles (AIChE) 22
8 Sample Calculations for Projectiles 23
9 Distances for Vapor Cloud Explosions of Four Liquefied Gases for
Instantaneous Releases of 1,000 to 180,000 Pounds Determined
Using Fraction Flashed, WHAZAN Stand-Alone Model 31
10 Quantity Released Versus Distance for Vapor Cloud Explosions,
Instantaneous Releases 32
11 Comparison of Explosion Data and WHAZAN Results 34
12 Distances for Vapor Cloud Fires from Evaporation Pools from
Instantaneous Releases, Moderate Meteorology, Determined Using WHAZAN 37
13 Quantity Released Versus Distance for Vapor Cloud Fires, Instantaneous
Releases, Moderate Meteorology 41
14 Distances for BLEVES for Instantaneous Releases of 1,000 to 180,000 Pounds .
Determined Using WHAZAN 43
15 Quantity Released Versus Distance for BLEVES, Instantaneous Releases 46
16 Distances for Pool Fires for Instantaneous Releases of 1,000 to 180,000
Pounds Determined Using WHAZAN 48
17 Quantity Released Versus Distance for Pool Fires, Instantaneous Releases 51
18 Distances for Jet Fires for Prolonged Releases of 10,000 Pounds Determined
Using WHAZAN 53
19 Flammable Gases 54
20 Flammable, Volatile Liquids : ;. 56
C-l Human Injury Criteria (Includes Injury from Flying Glass and Direct Overpressure
Effects) C-2
C-2 Property Damage Criteria C-4
-------
vu
LIST OF EXHIBITS
Exhibit Page
C-3 Lethality Curves C-7
C-4 Lethality and Injury Curves C-7
C-5 Air Blast Criteria for Personnel Standing in the Open C-8
C-6 Hazard Criteria Comparison for Vapor Cloud Explosions for Instantaneous
Releases of 100,000 Pound C-10
»
C-7 Hazard Criteria Comparison for Vapor Cloud Fires for Evaporating Pools
From Instantaneous Releases, Moderate Meteorology C-ll
C-8 Human Injury Criteria (Thermal Radiation Effects on Bare Skin) C-13
C-9 Fatality from Heat Exposure C-14
C-10 Hazard Criteria Comparison for BLEVEs, Instantaneous Releases C-16
C-ll Hazard Criteria Comparison for Pool Fires, Instantaneous Releases C-17
D-l Results of Modeling by All Methods D-2
E-l Temperatures and Pressures Used in Scenarios E-2
E-2 Chemical Data Used for Models E-3
E-3 WHAZAN Modeling Data for Instantaneous Releases, Moderate Meteorology '.... E-5
E-4 WHAZAN Modeling Data for Prolonged Releases E-7
E-5 WHAZAN Modeling Data for Instantaneous Releases, Worst Case Meteorology E-8
E-6 ARCHIE Modeling Data E-9
F-l Fire and Explosion Events Compared to All Events in Acute Hazardous
Events Data Base F-3
F-2 Flammable Gases and Liquids Reported in In-Plant Fires or Explosions F-4
F-3 Flammable Chemicals Involved in Fires and Explosions F-6
F-4 Summary of Explosion Accidents Involving Flammable Chemicals
in the ARIP Data Base F-7
F-5 Information About Explosion Accidents Involving Flammable
Chemicals in the OSHA Data Base F-10
F-6 Information About Explosion Accidents Involving Flammable Chemicals in
the M & M Data Base F-14
-------
EXECUTIVE SUMMARY
This report assesses the potential consequences of accidents involving flammable chemicals to
support the evaluation of whether such chemicals may warrant addition to the list of extremely hazardous
substances (EHSs) under section 302 of Title III of the Superfund Amendments and Reauthorization Act
(SARA). EPA's analysis included identification and evaluation of existing listing and classification
systems, along with any applicable criteria; review of existing regulations and codes dealing with flammable
materials; analysis of histories of accidents involving flammable substances; and modeling potential
consequences of fires and explosions of flammable substances.
EPA determined that most classification systems for flammable substances are based on
physical/chemical properties, usually flash point and boiling point The types of flammable substances that
are generally classified as most hazardous are flammable gases and volatile flammable liquids, with low
flash points and low boiling points. Evaluations of the severity of accidents that could be caused by
various flammable substances are not directly considered in the classification systems.
A review of accident history indicates that flammable substances have been involved in many
accidents, and, in many cases, fires and explosions of flammable substances have caused deaths and
injuries. Accidents involving flammable substances may lead to vapor cloud explosions, vapor cloud fires,
boiling liquid expanding vapor explosions (BLEVEs), pool fires, and jet fires, depending on the type of
substance involved and the circumstances of the accident
Vapor cloud explosions produce blast waves that potentially can cause offsite damage and kill or
injure people. EPA reviewed the effects of blast wave overpressures to determine the level that has the
potential to cause death or injury. High overpressure levels can cause death or injury as a direct result of
an explosion; such effects generally occur close to the site of an explosion. EPA's analysis of the literature
indicates that people also could be killed or injured because of indirect effects of the blast (e.g., collapse of
buildings, flying glass or debris); these effects could occur farther from the site of the blast. A vapor cloud
may burn without exploding; the effects of such a vapor cloud fire are limited primarily to the area covered
by the burning cloud. The primary hazard of BLEVEs, pool fires, and jet fires is thermal radiation; the
potential effects of thermal radiation generally do not extend for as great a distance as those of blast
waves. In addition, the effects of thermal radiation are related to duration of exposure; people exposed at
some distance from a fire would likely be able to escape. BLEVEs, which generally involve rupture of a
container, can cause container fragments to be thrown substantial distances; such fragments have the
potential to cause damage and injury. Fragments and debris may also be thrown out as a result of the
blast from a vapor cloud explosion.
The probability of occurrence of vapor cloud explosions appears to be rather low, based on
analysis of the literature. EPA reviewed factors that may affect the probability of occurrence of a vapor
cloud explosion, including the quantity of flammable vapor in a cloud, the presence of obstacles or partial
confinement, and the type of ignition source. Analysis of accidents indicates that vapor cloud explosions
are less likely when the quantity in the cloud is less than 10,000 pounds. It is generally thought that some
type of obstruction or confinement enhances the probability that a vapor cloud explosion, rather than a
vapor cloud fire, will occur. A high energy ignition source also contributes to the probability of
occurrence of a vapor cloud explosion.
EPA carried out consequence modeling for fires and explosions of a number of flammable
substances, using several PC-based modeling systems and hand calculation methods. Modeling results were
used to estimate the greatest distance at which people potentially could be killed or injured by explosions
or fires of flammable gases or liquids. The modeling indicated that, for a given quantity of a flammable
chemical, vapor cloud explosions may have the greatest potential for offsite consequences. This result is
consistent with information presented in the literature. Modeling indicated that BLEVEs may also,, in
some cases, have the potential for offsite consequences. Modeling results for pool fires and jet fires
-------
indicated that thermal radiation effects extend for much shorter distances than for BLEVEs, for a given
quantity of chemical.
Additional modeling was carried out based on actual vapor cloud explosions, and the results were
compared with the results of the actual incidents. In general, it was found that the modeling results were
in reasonable agreement with the results of the incidents. It was noted, however, that the specific
circumstances surrounding an accident may have a significant effects on the severity and range of
consequences of the accident. Modeling cannot take all circumstances and conditions into account.
Based on modeling and analysis of the literature, flammable gases and volatile flammable liquids
appear to be the flammable substances of most concern, because they may readily form vapor clouds, with
the potential for damaging vapor cloud explosions. EPA identified a number of such substances of
concern. The analysis carried out by EPA for this report was intended to provide general background on
the hazards of flammable gases and liquids. The modeling results and accident data illustrate and compare
the consequences of vapor cloud explosions, vapor cloud fires, BLEVEs, and pool fires. This analysis does
not provide a basis for determining the hazard posed by any flammable chemical in a specific situation.
-------
FLAMMABLE GASES AND LIQUIDS AND THEIR HAZARDS
1.0 Introduction
When the Emergency Planning and Community Right to Know Act (EPCRA), also known as Title
III of the Superfund Amendments and Reauthorization Act (SARA), was enacted in 1986, a list of
extremely hazardous substances (EHSs) (formerly the list of Acutely Toxic Chemicals published in
November of 1985) was published under Section 302 of Title III. Section 302 requires a facility that has
an EHS above its threshold planning quantity (TPQ) to notify the State Emergency Response Commission
(SERC) that it is subject to the requirements for emergency planning.
The original list of EHSs was created using only acute toxicity criteria. The purpose of the list was
to identify those substances that, if accidentally released, could cause death or serious irreversible health
effects from toxicity off-site after a short exposure. However, toxicity is not the only hazard posed by
chemicals. SARA states .that the Administrator may revise the list, but such revisions "shall take into
account the toxicity, reactivity, volatility, dispersibility, combustibility, or flammability of a substance."
The purpose of this report is to assess the potential consequences of accidents involving flammable
chemicals. A particular emphasis is placed on assessing the impacts of accidents on communities
neighboring industrial facilities.
The general approach taken in assessing the consequences of potential accidents involving
flammable chemicals was to identify and evaluate: 1) existing listing and classification systems, along with
any applicable criteria; 2) existing regulations and codes dealing with flammable materials; and 3) histories
of accidents involving flammable materials. EPA's evaluation is discussed in Sections 3 and 4.
None of the above supplied sufficient information to serve EPA's specific needs. It was, therefore,
decided to model the potential consequences from accident scenarios involving flammable chemicals to
establish physical and chemical properties that may be indicative of the hazards associated with these
chemicals. The results of the consequence modeling were then analyzed. The modeling is discussed in
Sections 5 through 9. EPA's analysis of the modeling results is discussed in Section 10.
2.0 Background
EPA originally developed the list of EHSs as part of the voluntary Chemical Emergency
Preparedness Program (CEPP) which was designed to increase public awareness of chemical hazards in
communities and to focus effort on emergency planning. EPA believed that communities needed a starting
point and intended that the list draw attention to the substances and facilities that pose the most
immediate concern based on toxicity from an emergency planning and response perspective. EPA
recognized and emphasized that there are tens of thousands of compounds and mixtures in commerce that
may pose a hazard under specific circumstances and that this list addressed lethality and serious
irreversible health effects associated with acute toxicity. The Agency chose lethality because it represents
the most immediate concern in an emergency situation.
Several commenters on the rulemaking for the EHS list noted that physical/chemical hazards such
as flammability or explosivity may lead to lethal consequences. In an official inquiry to the Agency in
1987, Senator Frank Lautenburg also noted concern for hazards other than toxicity and asked that EPA
focus on other hazards. The Agency agreed and noted its intent to evaluate hazards other than toxicity in
the future.
-------
-2-
A Physical/Chemical Criteria Workgroup (P/C Workgroup) was formed at EPA to address the
concerns raised by Senator Lautenburg and others and to focus on the problem of identifying materials
which are hazardous based upon their physical/chemical properties and could thereby be candidate EHSs.
The Workgroup initially considered the hazards referred to in the Act and determined that they fell into
two general categories: hazards related to toxic properties and hazards related to physical/chemical
properties. The characteristics of reactivity, volatility, dispersibility, combustibility, and flammability,
mentioned in the Act, are related to a variety of physical/chemical properties; the Workgroup chose to
focus on the physical/chemical properties that lead to the greatest potential hazards to people. Hazards
associated with physical/chemical properties are primarily based on two phenomena that can cause serious
personal injury or death. These phenomena are the result of the release of energy from highly reactive or
flammable chemicals and are identified as overpressures from blast waves and thermal ridiation from fires.
Overpressures result from nearly instantaneous energy release, or detonation, while thermal energy is
released during combustion, which occurs more slowly. Since explosive chemicals are highly reactive
substances that can detonate and create overpressures, and flammable chemicals can burn and produce
thermal radiation, the Workgroup focused on explosives and flammable chemicals as chemicals of concern.
The P/C Workgroup elected first to investigate what substances might have been identified and what
criteria might already be in use by other organizations to characterize hazardous chemicals and to
determine whether these criteria might meet EPA's needs. Criteria were investigated from many
organizations including the National Fire Protection Association (NFPA), the United States Department
of Transportation (DOT), the Occupational Safety and Health Administration (OSHA), and the European
Economic Community (EEC).
The workgroup initially sought to evaluate and possibly use the criteria from these other
organizations to develop options for adding to the list of EHSs chemicals that are flammable or explosive.
The criteria from other organizations did not appear to be based on the consequences of an accident in
terms of exposure to the community but rather on the consequences that could occur within the facility or
during shipment of bulk materials or during fire-fighting. For example, an NFPA flammability rating of 4
is assigned to materials that will burn readily and that disperse readily or vaporize rapidly or completely at
ambient conditions. This is important to fire-fighters approaching a fire involving a flammable material
but gives little indication of the impact on a community from an accident that takes place inside an
industrial facility.
The P/C Workgroup determined that an evaluation of the impacts of accidents involving substances
that may be hazardous to the community because of certain physical/chemical properties was necessary.
Such an evaluation would help to develop the criteria that would identify chemicals that should be
considered extremely hazardous and therefore of concern for community emergency response planning
efforts. One approach was to model accidents involving flammable chemicals and to analyze the resulting
data in order to establish a correlation between the severity of accident consequences to the community
and physical/chemical properties related to the consequences.
This approach is similar to the approach taken to establish the TPQs for the Section 302 EHSs
listed based on toxicity criteria. These TPQs were established by assuming an accident scenario, a loss of
containment of a specific toxic chemical, and then estimating the dispersion potential of each chemical.
An index value based on the dispersion potential and toxicity was used to rank the chemicals, and
chemicals were assigned to TPQ categories according to their ranking.
In its analysis of toxic substances under SARA section 302, EPA used a 100-meter fenceline
distance to provide a guideline for community emergency planners to use in setting priorities for planning
for hazards in the community. The 100-meter fenceline distance has also been used by the Occupational
Safety and Health Administration (OSHA) in its analysis for thresholds for chemicals listed in its Process
Safety Management Standard and by the state of Delaware for thresholds under its prevention regulations.
-------
-3-
EPA recognizes that this distance may not be appropriate for protection of emergency responders. There
is no absolute distance that would guarantee the safety of first responders in every situation.
The workgroup elected to segregate the chemicals into separate categories for flammables,
explosives, and reactives for purposes of analysis and possible regulation even though it was recognized
that there may be overlap of the consequences; e.g., a detonation of a flammable material may yield the
same results as an explosion of a commercial explosive. However, the workgroup wanted to determine
what parameters distinguish an extremely hazardous flammable substance from all other flammables,
extremely hazardous explosives from all other explosives, and so on. This report focuses on flammable
gases and liquids. A separate document has been developed to address commercial explosives and their
hazards.
3.0 Existing Classification Systems, Regulations, and Codes
There are many organizations that have developed lists, definitions, and classifications related to
flammable chemicals, including DOT, NFPA, EEC, OSHA, and EPA. Some organizations establish their
classifications with qualitative descriptions, but most classifications are based on physical/chemical
properties, usually flash point and boiling point. None of these classification systems appeared to be
primarily based on potential hazards to the community from incidents involving releases of flammable
chemicals from fixed facilities; therefore EPA decided that none of these systems was appropriate or
specifically applicable to EPA's regulatory needs. The systems examined are described briefly in the
following sections; more detail is presented in Appendix A.
EPA intends to pursue harmonization of regulatory activity among various agencies to the extent
possible, recognizing the different purposes of the regulations under different agencies. Regulation under
section 302 of SARA Title III is intended to provide information to community planners on substances at
fixed sites that may be hazardous to the community. As noted below, the regulations and classification
systems used by other agencies and organizations have other purposes.
3.1 DOT Classifications
DOT classifies flammable materials for shipping purposes under 49 CFR 172. This regulation lists
materials DOT regards as hazardous for purposes of transportation and prescribes requirements for
shipping papers, package marking, labeling, vehicle placarding, and types of containers and safety devices
that must be used to transport a flammable material. Details of the DOT classifications for flammable
materials are found in Appendix A.
The DOT hazard classifications are based on maintaining safety during transport, given the range of
ambient conditions possible. The consequences in terms of "first response" to accidents involving these
materials are addressed in DOTs Emergency Response Guidebook. This guide details generic isolation
distances, recommended fire-fighting techniques, and other initial emergency actions for incidents involving
these substances. However, DOT focuses on transportation safety rather than hazards from fixed sites.
DOT hazard classes and packing groups for flammable substances are generally based on flash point
and boiling point ranges, as described in Appendix A. However, for substances that are flammable and
toxic, the packing group may be modified to reflect the additional hazard of toxicity; therefore, the DOT
classifications of such substances may not provide a measure of their flammability.
3.2 NFPA Classification Systems
The NFPA has a flammability rating scale of 0 to 4, where 0 represents the lowest degree of
flammability, for purposes of fire-fighting and fire prevention. The ratings are usually noted on Material
-------
Safety Data Sheets (MSDSs) and are included in a standard that applies to facilities for manufacturing,
storage, or use of hazardous materials. Approximately 1,500 chemicals have been assigned NFPA ratings.
The classifications are designed to give "a general idea of the inherent hazards of any material and the
order of severity of these hazards as they relate to fire prevention, exposure and control" (NFPA 1984).
NFPA has also developed a code (NFPA 30, Flammable and Combustible Liquids Code)
concerning flammable and combustible liquids which is pan of the National Fire Code and is specifically
intended to be referenced by public authorities in laws, ordinances, regulations, and administrative orders.
It was originally written as a model municipal ordinance for storing, handling, and using flammable liquids
at fixed facilities. This code includes classifications based on flash point and boiling point. The NFPA
flammability ratings and classifications from the NFPA code are discussed further in Appendix A.
3.3 EEC Classification Scheme
The EEC published a Council Directive on June 24,1982, "concerned with the prevention of major
accidents which might result from certain industrial activities and with the limitation of their consequences
for man and the environment" (EEC 1982). The directive includes a list of toxic and explosive chemicals
and several categories of flammable chemicals; the flammable categories are defined by "Indicative
Criteria." Facilities that have listed chemicals in quantities greater than a specified threshold amount are
required to report and meet a variety of conditions including performing hazards analyses and preparing
emergency plans. The thresholds for the flammable categories are 200 metric tons (440,000 pounds) for
flammable gases and flammable liquids under hazardous processing conditions, and 50,000 metric tons (1.1
million pounds) for other highly flammable liquids. The EEC Directive does not include a list of
chemicals that fall into the various flammable categories. The indicative criteria related to flammability
are discussed in Appendix A.
3.4 Regulations and Codes Applicable to Flammable Chemicals
Several agencies regulate flammable chemicals. OSHA regulates flammable and combustible liquids
(29 CFR 1910.106), with handling and storage requirements and specifications for storage tanks. Specific
requirements are included for bulk plants, service stations, processing plants, refineries, chemical plants,
and distilleries. In addition, there are specific regulations dealing with acetylene (29 CFR 1910.101),
hydrogen (29 CFR 1910.102), and liquefied petroleum gases (29 CFR 1910.110).
Under the Clean Air Act Amendments of 1990, OSHA was required to promulgate regulations
intended to prevent accidental releases of chemicals which could pose a threat to employees. OSHA has
published a standard, as described below. EPA is required to develop regulations for protection of the
public, as discussed below.
OSHA has developed a chemical process safety management standard (57 FR 6356, February 24,
1992) that applies to any process involving a highly hazardous chemical at or above a specified threshold
quantity. The requirements of this standard also apply to processes that involve flammable gases and
flammable liquids in quantities of 10,000 pounds or more. Liquids stored below their boiling points under
ambient pressure, without refrigeration, are exempt from the requirements. The OSHA Process Safety
Management Standard is intended to protect employees by preventing or minimizing the consequences of
catastrophic releases of toxic, reactive, flammable, or explosive chemicals. The requirements include
development of a compilation of written process safety information; process hazards analysis, carried out
by a team with expertise in engineering and process operations; development of written operating
procedures for processes involving highly hazardous chemicals; employee training; pre-startup safety
reviews for new and significantly-modified facilities; maintenance of mechanical integrity of critical
equipment; and establishment of procedures for management of changes to process chemicals, technology,
-------
-5-
equipment, and procedures; investigation of incidents; emergency action plans; and compliance and safety
audits to ensure programs are in-place and operating properly.
Section 112(r) of the Clean Air Act (CAA) requires EPA to develop a list of at least 100 regulated
substances that, when released, can cause death, injury, or serious adverse effects to human health or the
environment. EPA proposed a rule listing regulated substances and associated thresholds on January 19,
1993 (85 FR 5102). The proposed rule includes a list of 68 flammable gases and volatile flammable liquids
at a threshold of 10,000 pounds. The listed flammable chemicals meet the flash point and boiling point
criteria for NFPA 4 (see above) and are considered the substances with the highest potential for
involvement in vapor cloud explosions (see Section 6.1). In addition, substances proposed for regulation
include 100 toxic chemicals, as well as commercial explosives as defined by DOT in Division* 1.1. Facilities
that use these chemicals above threshold quantities developed by EPA need to comply with new CAA
regulations on release prevention, detection, and emergency response. One accident prevention provision
of the CAA mandates the development of regulations requiring facilities to prepare and implement risk
management plans. EPA is currently developing risk management plan regulations.
Flammable and combustible liquids may also be regulated by state or local governments; e.g., NFPA
30, Flammable and Combustible Liquids Code (see section 3.2), may be adopted or incorporated into law.
DOT regulations include shipping requirements for flammable materials (49 CFR 172 and 173). All of
these regulations are highly specific and deal with containers, pipes, and storage, handling, or transfer
conditions in specified circumstances; however, none is directly applicable for EPA's purposes.
Gasoline is a flammable chemical of particular interest because of the high volumes in commerce
and the many locations where it may be found. Therefore, EPA investigated the regulations for flammable
chemicals with additional focus on gasoline. Gasoline is not regulated as a special case under any of the
regulations or codes mentioned above. DOTs requirements for gasoline are similar to those for other
liquids with similar properties, as are NFPA's recommendations for gasoline. OSHA's regulations for
service stations do not mention gasoline specifically, but refer only to flammable liquids. These
regulations cover the storage and dispensing of flammable liquids, electrical and heating equipment in
areas with flammable liquids, waste handling, and fire extinguishers at service stations. Further
information on the OSHA regulations for flammable liquids may be found in Appendix A. EPA's review
of current regulations indicates gasoline is not considered a special case.
Fuels are not included as hazardous substances under CERCLA; "petroleum, including crude oil or
any fraction thereof, ...natural gas, natural gas liquids, liquefied natural gas or synthetic gas usable for fuel
(or mixtures of natural gas and such synthetic gas)" are not included in the definition of hazardous
substances (40 CFR 300.6).
4.0 Past Fire and Explosion Incidents Involving Flammable Chemicals
EPA decided to review accidents involving flammable chemicals that have occurred in the past and
their impact on the community to determine whether trends could be identified that would aid in
identifying flammable chemicals that may be particularly hazardous to the community. The results of
EPA's review are presented in this section.
4.1 Acute Hazardous Events Data Base
The Acute Hazardous Events Data Base (AHE/DB) was developed by EPA to provide a historical
perspective on accidents that involve releases of hazardous substances in the United States (U.S. EPA
1988). The 1988 version of the data base includes data on 5,827 incidents that occurred between the 1960s
and 1987 (the majority of the incidents appear to have occurred in the 1980s). Information in the data
base is from the National Response Center, press reports, and several state offices. This data base was
-------
-6-
reviewed to identify accidents specifically involving flammable chemicals and evaluate the consequences of
these accidents, if possible, in terms of physical/chemical characteristics and the specific hazards posed to
the surrounding community.
The AHE/DB indicates that flammable chemicals have been involved in numerous accidents that
resulted in fires and explosions, and these accidents led to many deaths and injuries. However, the data
are not sufficiently specific to serve as a basis for identifying chemicals that may be extremely hazardous
based on their physical/chemical properties.
Fire and explosion incidents account for about 13 percent of the total incidents in the AHE/DB,
but they account for nearly 92 percent of the deaths and about 36 percent of the injuries' reported.
However, the number of these deaths and injuries that occurred on-site and the number that occurred off-
site cannot be differentiated in the data base; therefore, the degree of impact these incidents had on the
surrounding community cannot be accurately determined. Additional details and analyses using the data
from the AHE/DB may be found in Appendix F.
4.2 Accidental Release Information Program
EPA's Accidental Release Information Program (ARIP) data base contains 2,200 records (as of
June, 1992) with questionnaire information from facilities on the causes and consequences of accidental
releases and on release prevention procedures and equipment. Flammable chemicals included in the ARIP
data base are either CERCLA hazardous substances or EHSs. Facilities must complete the ARIP
questionnaire if the release meets at least one of the following criteria:
1) The quantity released was above a certain multiple of the CERCLA Reportable Quantity
(RQ);
2) The release resulted in a death or injury;
3) The release was one in a trend of frequent releases from the same facility; or
4) The release involved a chemical listed by EPA as an EHS.
Of the data bases reviewed, only the ARIP data base maintains information on off-site deaths,
injuries, or evacuations. Based on analysis performed on 2,200 ARIP records, six events have been listed
as explosive incidents involving flammable chemicals. None of the six caused a public death or injury.
One ARIP release and explosion of propylene caused two worker injuries and evacuation of ISO members
of the public. In one incident, explosion of silane gas resulted in the deaths of two facility workers.
Appendix F provides additional information on these six releases.
4.3 OSHA Data Base
The OSHA data base contains records of OSHA accident inspections. These inspections were
conducted in response to a worker death, at least three worker injuries, or reporting of workplace hazards.
The most recent and complete data base records for OSHA inspections (fiscal year 1990) were examined
for this analysis. Analysis of the data indicate, that 32 incidents involving flammable chemicals resulted in
25 deaths of workers and 43 injuries to workers (OSHA 1990). See Appendix F for more.information on
these releases.
-------
-7-
4.4 Fatal Hazardous Materials Accidents Database
The Fatal Hazardous Materials Accidents Database, developed by Resources for the Future,
contains information on fatal accidents which involved the release of hazardous materials. Data were
obtained from encyclopedias, almanacs, books, reports, articles, newspapers and computer files (including
AHE/DB). The database contains reported accidents which occurred during industrial production, storage,
handling and transportation of hazardous materials. Only those accidents involving at least one reported
death for U.S. accidents and five reported deaths for accidents outside the U.S. are included. Every such
event between 1945 and 1991 is in the database, with the following exceptions: mining or mineral
extraction accidents; handling, transportation, and storage of munitions, fireworks, and manufactured
explosives (manufacture of such products was counted); and transmission and distribution of natural gas
(i.e., pipelines).
There are 1,068 accidents reported in the database, of which 758 occurred in the U.S. Fifty-six
percent of these fatal accidents involved a fire or explosion of a flammable material. Explosions of
flammable substances accounted for 412 accidents, and fires (excluding fires involving explosives)
accounted for 188 accidents in the database. Approximately 14 to 25 thousand injuries, and eight to 12
thousand deaths were associated with these accidents. The database does not distinguish between on-site
and off-site injuries and fatalities. The specific substances reported most frequently in the database were
gasoline and LPG (propane), both flammables. The types of reported fires and explosions of flammable
substances in the database included 55 unconfined vapor cloud explosions, 43 confined vapor explosions,
48 fireballs, 22 BLEVEs, and 11 detonations of flammable substances. In a number of'cases, two of these
types of events were reported for the same incident (e.g., unconfined vapor cloud explosions and BLEVEs
were reported in three incidents, unconfined vapor cloud explosions and fireballs were reported in seven
incidents).
4.5 Major Hazard Incident Data Service (MHIDAS)
The Major Hazard Incident Data Service (MHIDAS) is a world-wide databank maintained by the
Safety and Reliability Business (SRD) of AEA Technology and the UK Health and Safety Executive. It is
continually updated, and records and analyzes incidents involving hazardous materials that resulted in or
had the potential to produce an off-site impact. Off-site impact includes incidents which involved
evacuation and those which could, but for mitigating circumstances, have led to evacuation, casualties, or
damage to property or the environment.
MHIDAS currently contains over 6,500 records with information on incidents involving hazardous
substances. Nearly 5,000 of these records involve fires or explosions or the threat of fire (e.g., a flammable
material was involved in the incident. Types of fire or explosion incidents, their frequency in the database,
and the number of records in which deaths, injuries, or evacuations were reported for flammable
substances shown in the table on the next page.
As the table shows, it appears that a somewhat higher percentage of unconfined vapor cloud
explosion records in the database include reports of deaths, injuries, or evacuations, with an even higher
percentage for deaths, than the other types of fires and explosions. However, it should be noted that the
numbers reported above do not precisely reflect the number of incidents reported. Because reported
incidents may have involved multiple fires or explosions, the numbers presented reflect some overlap. In
addition, if several substances were involved in an incident, there may be multiple records for that incident
in the database.
-------
-8-
. Type of Incident
Confined explosions involving flammable
substances
Unconfined vapor cloud explosions
BLEVEs involving flammable substances
Fireballs
Vapor or flash fires
Pool fires
Jet fires
Number of
Records
197
110
111
98
96
113
15
Number with
Deaths, Injuries,
or Evacuations
136 (69 percent)
89(81 percent)
84 (76 percent)
72 (73 percent)
72 (75 percent)
73 (65 percent)
8 (53 percent)
Number with
Deaths
72 (37 percent)
63 (57 percent)
4l (37 percent)
34 (35 percent)
51 (53 percent)
39 (35 percent)
5 (33 percent)
4.6 M & M Protection Consultants Data Base
The M & M Protection Consultants document entitled Large Property Damage Losses in the
Hydrocarbon-Chemical Industries - A Thirty-Year Review describes 100 incidents involving large property
damage losses. Offshore and marine transportation incidents are excluded. Because the data base focuses
on hydrocarbon-chemical incidents, it covers many accidents involving flammable chemicals. Most of the
incidents involve fires or explosions. Vapor cloud explosions account for 44 of the 100 incidents in the M
& M data base. The document notes, "The vapor cloud explosion ... has become established as the
mechanism most frequently leading to catastrophic losses in the hydrocarbon.processing and chemical
industries" (M & M 1990). Vapor cloud explosions are discussed more fully in Section 6.1.
The M & M data base is particularly valuable in providing information on the consequences of
vapor cloud explosions on the community nearby a facility. Spanning thirty years, the data include
descriptions of explosion and fire events, the causes of the property loss, and the estimated costs of
property damages. Information specifically related to consequence analysis includes quantity released,
temperature and pressure of chemical released, overpressures created by an explosion, projectile distances,
vapor cloud size, flashback distances, and property damage. Unfortunately, data on death and injury are
not available in the data base as provided. Appendix F provides additional information on incidents in the
M & M data base.
Of the data bases consulted, the M & M data base provides the most complete information on the
circumstances and the off-site consequences of flammable chemical incidents, particularly on vapor cloud
explosions. However, the small number of incidents over a thirty year period is not sufficient for EPA to
identify or classify flammable chemicals that pose hazards to the community.
5.0 Accident Consequence Modeling
Because current classification methods and regulatory approaches used by various organizations
(e.g., DOT, NFPA, EEC, OSHA, etc.) are not oriented to addressing the hazards posed to the community
from flammable chemicals and because accident data also are not sufficient for identifying or classifying
the hazards of flammable chemicals, EPA decided to use modeling techniques to evaluate potential hazards
-------
-9-
to the community from incidents involving release of flammable liquids and gases from fixed sites. No
single hazards analysis model has been universally accepted; therefore, three PC-based models and several
methods involving manual calculations were used for a single release quantity to estimate consequences of
accidents involving flammable liquids and gases for comparison purposes. As will be discussed in the
following section, the WHAZAN model was selected for use in the accident consequence analysis for a
range of quantities and chemicals.
The remainder of this section provides an overview of various models available. Section 6 discusses
the different types of flammable chemical fires and explosions. Section 7 examines how certain
characteristics of flammable chemical accidents (e.g., confinement, delayed ignition) will qualitatively affect
the modeling results. Additionally, inputs for consequence models are discussed. Section 8*examines the
hazards criteria used for evaluating the consequences of different types of flammable chemical incidents.
Section 9 provides the results of the modeling runs, including WHAZAN results using data from actual
vapor cloud explosions and a comparison of the WHAZAN results with the actual results of the incidents.
Section 10 details EPA's findings.
EPA did not attempt to evaluate the models used or to estimate uncertainties in modeling results.
The results obtained using the various methods are presented for comparison purposes in Appendix D. In
general, the various methods show reasonable agreement in predicting consequences for various accident
scenarios and hazard criteria levels. Additional information on each of the models is presented in
Appendix B. Assumptions used in the modeling and the physical/chemical data required as input are
summarized in Appendix E.
5.1 WHAZAN Methodology
The commercially available World Bank Hazard Analysis (WHAZAN) computer system is a group
of computer programs capable of performing quick estimates of the possible hazardous consequences of
accidental releases of toxic and/or flammable gases or liquids. It was developed by Technica International
Ltd. in collaboration with the World Bank. These programs comprise several models which can calculate
the consequences of different types of chemical accidents. The models derive their inputs from a data base
which contains relevant properties of some hazardous chemicals. The user can add chemicals and
properties to the data base. A user of WHAZAN can either run one of the models individually (stand-
alone models) for calculating the hazardous consequences from a specific type of event (e.g., vapor cloud
explosion), or can link two or more models in a way such that outputs from one model can be used
automatically as inputs for another model (linked models). The models provide information about four
key areas: outflow of the chemical from its container, behavior immediately after release, dispersion after
release, and fires and explosions. The WHAZAN system is extensively used by large international
chemical companies and government agencies.
WHAZAN makes certain assumptions regarding atmospheric behavior and accidental release rates
of various gases and liquids. Comparison of WHAZAN results with the results using other methods
showed reasonable agreement. Also, WHAZAN results presented in section 9.1 largely agree with
consequence data from actual vapor cloud explosions. Therefore, WHAZAN was used as the primary
method of consequence analysis for flammable chemicals. More details on the WHAZAN model and its
theoretical basis are found in Appendix B.
5.2 ARCHIE Methodology
Accident hazard assessment and consequence analysis procedures have been incorporated into a
system developed for the Federal Emergency Management Agency (FEMA), DOT, and EPA, entitled
Automated Resource for Chemical Hazard Incident Evaluation (ARCHIE). The primary purpose of
ARCHIE is to provide emergency personnel with several integrated estimation methods that may be used
-------
-10-
to assess the vapor dispersion, fire, and explosion impacts associated with episodic discharges of hazardous
materials into the environment. The program also is intended to facilitate a better understanding of the
nature and sequence of events that may follow an accident and the resulting consequences.
The core of the ARCHIE computer program is a set of hazard assessment models that can be
sequentially utilized to evaluate consequences of potential discharges of hazardous material and thereby
assist in the development of emergency plans. ARCHIE can help emergency planning personnel
understand: 1) the nature and magnitude of hazards posing a threat to their jurisdiction; 2) the sequence
of events that must take place for these threats to be realized; and ultimately 3) the nature of response
actions that may be necessary to mitigate adverse impacts upon the public and its property in the event of
an emergency. »
The system contains numerous models for evaluating different types of incidents and their
consequences, e.g., pool fires, unconfined vapor cloud explosions, and fireballs. More details about the
system and its theoretical basis are presented in Appendix B.
5.3 The Tellow Book" Methodology
The Bureau of Industrial Safety (TNO) of the Netherlands has issued a report, commissioned by the
Directorate-General of Labour, entitled "Methods for the Calculation of the Effects of the Escape of
Dangerous Material" (March 9,1980), also known as the "Yellow Book." Equations and calculation
methods for estimating the effects of several types of accidents involving flammable liquids and gases are
presented. Some of these methods were used as the basis for models in WHAZAN. The Yellow Book
includes calculations for pool fires, vapor cloud fires, and vapor cloud explosions for gases, and pool fires
and vapor cloud fires for liquids. For gases, two different methodologies are used for vapor cloud fires:
the flashed fraction of each gas is modeled as an instantaneous release; the non-flashed fraction is modeled
as a continuous release from an evaporating pool. In both cases, dispersion as a neutrally buoyant cloud is
assumed. The flashed fraction of each gas is used to calculate the vapor cloud explosion results. Since
there is no calculated flashed fraction for liquids, no vapor cloud explosion results appear for liquids.
Appendix B presents details of the calculations that were used in this analysis.
5.4 AIChE-Sponsored Course Materials
The American Institute of Chemical Engineers (AIChE) sponsored course material titled, "Methods
for Calculation of Fire and Explosions Hazards" (1987), presents equations for calculating the effects of
several types of accidents involving flammable liquids and gases. This material is a compendium of
methods and algorithms dealing with fires and explosions, not a straightforward approach to determining
consequences of various accident scenarios. The equations used in this analysis and details of the
methodology are presented in Appendix B.
5.5 The "Green Book" Methodology
EPA, in conjunction with FEMA and DOT, developed the "Technical Guidance for Hazards
Analysis" (1987) to provide guidance to local emergency planning committees (LEPCs), local emergency
agencies, and community groups in assessing hazards related to potential airborne releases of extremely
hazardous substances listed because of their toxicity under section 302 of SARA Title HI. This guidance,
known as the "Green Book," provides methods for estimating chemical releases and dispersion distances.
These methods were applied to the airborne dispersion of vapor clouds of flammable liquids and gases to
estimate distances for vapor cloud fires. Appendix B contains details of these calculations.
-------
-11-
5.6 PHAST Methodology
PHAST (Process Hazard Analysis Software Tool) is a software package developed by Technica
International Ltd. and Rohm & Haas. Technica also developed WHAZAN for the World Bank. In fact,
PHAST uses the same algorithms as WHAZAN for predicting consequences. PHAST, like WHAZAN, is
designed to estimate the consequences of accidental releases of toxic or flammable liquids and gases.
Unlike WHAZAN, PHAST is a PC-based version of SAFETI (Software for Assessing Flammable,
Explosive, and Toxic Impact) a mainframe-based hazard analysis system owned by Technica. PHAST is
used by Technica to evaluate potential changes to SAFETI, and therefore represents one of the most
current hazard modeling packages. The PHAST package includes a data base of 56 commonly used
chemicals; additional chemicals may be entered. The PHAST system can be used for estimation of the
effects of BLEVEs, pool fires, jet flames, flash fires, and vapor cloud explosions.
There are some important differences between PHAST and WHAZAN. PHAST incorporates the
TECJET model, a proprietary dispersion model developed by Technica. TECJET provides algorithms
which simulate aerosols', overflows, and gas dispersion in PHAST. TECJET is a three dimensional
dispersion model which accounts for perturbation around the center line axis of a release caused by a
crosswind. WHAZAN does not have this capability. Unlike WHAZAN, the models in PHAST cannot be
run independently as stand-alone models, but are completely integrated to run in sequence. PHAST
includes algorithms which maintain strict thermodynamic control during the modeling. This prevents
unrealistic modeling runs such as calculating a vapor cloud explosion when thermodynamic conditions
dictate that the release will not form a vapor cloud.
5.7 Multi-Energy Model
A.C. van den Berg and others from TNO introduced the multi-energy concept for vapor cloud
explosions (van den Berg 1985; van den Berg et al. 1991). In contrast to the TNT equivalence concept
where the entire vapor cloud is treated as a potential source of a vapor cloud explosion, the multi-energy
method is based on the principle that only partially confined areas in the cloud are likely to contribute to
blast generation. Each such area is treated separately. The initial blast strength for each area is expressed
as a number ranging from one to ten, depending on degree of confinement and obstruction; a factor of ten
applies to a detonation. The total combustion energy involved and the initial blast strength are used to
derive a distribution of blast parameters in the vicinity of the gas explosion. A mathematical model has
been developed to solve equations describing convective and diffusive transport of flow. Small scale
experiments were carried out, and the results were compared with results from the model. A research
program on this method is still being carried out (van den Berg et al. 1991).
Some drawbacks of the multi-energy method have been noted. Lenoir and Davenport (1993)
suggest that many variables need further evaluation, including degree of confinement, initial blast strength,
and strength of ignition sources. In addition, the multi-energy method only applies to quiescent vapor
clouds, not to explosively dispersed clouds (Lenoir and Davenport 1993).
The multi-energy model appears to be more complex than the PC-based models and calculations
discussed in the previous sections. Results depend on site-specific factors such as obstructions present,
which are not considered in the other models; therefore, it would be difficult to compare results from this
model to results from the other models. Calculations based on this method cannot be carried out without
a computer program; EPA has not attempted to obtain such a program and has not included the multi-
energy method in this analysis.
-------
-12-
5.8 New Jersey Method for Fireballs
The New Jersey Department of Environmental Protection and Energy has developed a computation
method for calculating quantity-distance relationships for the thermal effects of fireballs from Boiling
Liquid Expanding Vapor Explosions (BLEVEs). New Jersey has developed a computer program,
"ACTOR Model Thermal Energy Analysis Subroutine," to carry out the calculations (NJDEPE 1991). A
review of this method indicated that it is similar to the WHAZAN and ARCHIE methods for estimating
effects of BLEVEs.
6.0 Types of Flammable Chemical Fires and Explosions
t
There are several types of flammable chemical incidents that may have consequences for the
community. Among the literature sources examined, there appears to be a general consensus regarding
the types of events resulting from releases of flammable substances. These include vapor cloud explosions,
vapor cloud fires, BLEVEs, pool fires, and jet fires, as discussed in the following sections. The hazards
posed to the public from these events are also described below, as well as factors (e.g., quantity released,
degree of confinement) that influence the probability of occurrence and the severity of the consequences of
the explosion or fire. The hazards of projectiles resulting from explosions are also discussed.
6.1 Vapor Cloud Explosions
Volatile materials can form vapor clouds when they are released in an uncontrolled manner from
containment vessels. Materials that are at elevated temperatures and/or pressures are particularly of
concern. Released volatile, flammable materials can flash directly into the vapor state and/or flow to the
ground and form a pool of liquid which subsequently evaporates to form a vapor cloud. After a vapor
cloud is formed, three things can happen: it can disperse and have no fire or explosion impact on
neighboring communities; it can ignite and burn (vapor cloud fire); or it can detonate or explode and
cause high pressure shock waves. According to Kletz (1977), unconfined vapor cloud explosions almost
always result from the release of flashing liquids (i.e., liquids under pressure at temperatures above their
boiling points). For a vapor cloud to ignite and burn or explode, its concentration in air must be within
its flammable limits.
Chemicals are vaporized much more rapidly by boiling than by evaporation; therefore, chemicals
with boiling points below ambient temperatures volatilize very rapidly after being spilled and are more
likely to generate high-mass clouds than chemicals that boil above ambient temperatures. Chemicals that
are gases at ambient temperatures therefore represent a greater explosion hazard than less volatile
chemicals with boiling points greater than ambient temperatures (liquids and solids). However, vapor
clouds may also be formed by higher-boiling liquids (e.g., cyclohexane) released at temperatures above
their boiling points.
An exemplary list of chemicals that are known to have been involved in vapor cloud explosions is
shown in Exhibit 1. Note that 12 of the IS chemicals listed are gases under ambient conditions.
Exhibit 2 summarizes information on a number of incidents that resulted in vapor cloud explosions.
These incidents are cited in later sections to illustrate the effects of various factors that may influence the
outcomes of incidents involving flammable chemicals. In cases where there were sufficient data, modeling
was carried out using reported data from selected incidents as model input The modeling effort and the
results obtained are discussed in Section 9.2.
-------
-13-
Exhibit 1
CHEMICALS KNOWN TO HAVE BEEN INVOLVED IN VAPOR CLOUD EXPLOSIONS
Chemical
Hydrogen
Methane/LNG
Ethylene
Propylene
Dimethyl Ether
Propane
Vinyl Chloride
Isobutane
Isobutylene
Butadiene
Elhylene Oxide
Ethyl Chloride
Gasoline
Hexane
Cyclohexane
References:
Boiline Point (°C1
-253
-161
-102
-46
-43
-42
-24
-14
-12
- 8
11
12
59
72
81
R.W. Prugh (1987)
J. Skarka (1987)
Exhibit 2
SAMPLE QUANTITIES INVOLVED IN VAPOR CLOUD EXPLOSIONS
Facility
Quebec, Canada
New Castle, DE
Texas City, TX
Norco, LA
Doe Run, KY
Flixborough, England
Lake Charles, LA
Chemical
styrene
propylene
hydrocarbons
hydrocarbons (propane)
ethylene oxide
cyclohexane
isobutane
Quantity in Cloud (pounds)
1,500
12,000 - 16,000
20,000
20,000
45,000
60,000
(spilled)
80,000
(spilled)
Source: M & M Data Base, 1990
-------
-14-
Vapor cloud explosions can be extremely destructive incidents that can have both on- and off-site
impacts. Usually, the high pressure or overpressure created by the explosion is responsible for the
' damage. In addition, projectiles (e.g., container fragments, debris from damaged buildings) can cause
damage and injury (see Section 6.6 for a discussion of projectiles).
6.1.1 Effect of Quantity on Probability of Vapor Cloud Explosions
According to Prugh (1987), the probability of a vapor cloud exploding appears to depend on the
weight of flammable vapor in the cloud (i.e., the quantity in the cloud above the lower flammability limit).
Prugh's analysis (based on actual vapor cloud explosion events) indicates that the probability of explosion
is 50 percent when the weight of flammable vapor is 90,000 kilograms (200,000 pounds)* the probability
decreases with decreasing weight. Based on an equation developed by Prugh, for vapor clouds containing
less than 10,000 pounds of flammable vapor, the probability of explosion becomes relatively low
(approximately 7 percent for 10,000 pounds and less than 1 percent for 1,000 pounds). Prugh's probability
analysis considered only the likelihood of a vapor cloud exploding following a release, not the likelihood
that a release might occur.
The American Petroleum Institute (API) has developed recommended process hazard management
practices (including such topics as safety reviews, hazards analyses, training, safe practices, and incident
investigation) for production and refining departments (API Recommended Practice 750). This document
has an appendix detailing the application of the recommended practice to facilities that have the potential
to release five tons (10,000 pounds) of gas or vapor in a period of a few minutes. The five-ton quantity
was chosen by API as a reasonable threshold on the basis of catastrophic potential and probability of
explosion. For a vapor cloud containing five tons of hydrocarbon, API considers the probability of
explosion to be about 5 percent, compared to less than 1 percent for one ton or less (API 1990).
Data from the M & M data base generally seem to support the 10,000-pound threshold for vapor
cloud explosions (see Exhibit 2). The explosion in Canada involving 1,500 pounds may be an anomaly
since it was a partially confined vapor cloud explosion. Vapor cloud explosions have occurred at quantities
as low as 20 kg, but the damage they have produced has been very localized and of little concern regarding
injuries and losses (Lewis 1980). Releases as great as 1,000 metric tons have been identified, but they have
usually produced fires or firestorms. The rate of discharge has a large influence on air mixing. Rates of
discharge lower than about 0.25 metric tons of fuel per minute do not seem to be a severe aerial explosion
risk (Lewis 1980), probably because the low rate of release is unlikely to lead to accumulation of a cloud
of vapor within the explosive limits.
6.1.2 Confinement and Mixing in Vapor Cloud Explosions
Presence of obstacles may confine the vapor cloud and consequently increase the concentration of
the flammable substance in the vapor cloud. Upon deflagration or detonation, this increased
concentration ensures flame acceleration and/or maximum overpressure achievement. In addition, the
presence of objects in the cloud's path appears to affect the likelihood of explosion. Wiekema (1984)
concludes that, "...when no sizable objects were in the cloud no explosions have been recorded, and that, in
fact, the presence of obstacles and confinement is a necessary condition for an explosion to occur." Also,
confined vapor cloud explosions or fires require less cloud mass than unconfined vapor cloud explosions or
fires. Lewis (1980) reports that "significant" blast effects have occurred under semi-confined conditions for
quantities as low as 100 kg (220 pounds); Lewis does not further describe the blast effects. One case in
the ARIP database involved confinement and subsequent explosion.of 1,000 pounds of propylene and
propane (ARIP 1991).
A study by Prugh (1987) concludes that the occurrence of a vapor cloud explosion rather than a
vapor cloud fire (both of which occur under similar circumstances) depends upon the location of the
-------
-15-
ignition source and the amount of turbulence or mixing in the cloud. Skarka (1987) concludes that only
the central zone of the cloud of vapors mixed with air inside the explosion limits can cause and contribute
to the explosion.
According to the TNO Yellow Book (1980), the composition of vapor clouds is not the same at
every point; there are concentration gradients and inhomogeneities. The effect of these concentration
gradients and inhomogenities on the flame acceleration mechanism of a detonation is not fully known.
According to van den Berg et al. (1991), the inhomogeneity of vapor clouds resulting from an accidental
release prevents the propagation of a detonation through the cloud in most cases; therefore, most vapor
cloud explosions may be assumed to be deflagrations (van den Berg et al. 1991).
6,13 Energy of Vapor Cloud Explosions
Many flammable gases have relatively high theoretical TNT equivalents compared to commercial
explosives, where the TNT equivalent is the relative weight of trinitrotoluene (TNT) that produces the
same effect as is produced at a given distance by the material under consideration. Commercial explosives
generally have TNT equivalents between roughly 0.3 and 2. The theoretical TNT equivalent for flammable
gases indicates the explosive potential, assuming the material explodes completely; calculated values
reported by Prugh (1987) range from 4.5 (for vinyl chloride) to 34 (for hydrogen). For most hydrocarbons,
the theoretical TNT equivalent is about 12 (Prugh 1987). However, in vapor cloud explosions, it appears
that only a fraction of the vapor in the cloud explodes. (In contrast, for commercial explosives, most of
the explosive material present is expected to be involved in the explosion.) For example, Prugh (1987)
reports that for vapor cloud explosions of hydrocarbons, the observed TNT equivalent was 1.0, compared
to the theoretical TNT equivalent of 12 (i.e., the actual explosive yield is eight percent of the theoretical
yield. Lees (1980) reports that the fraction of the heat of combustion used to produce the blast wave is
usually between one and 10 percent but that much higher explosion efficiencies have been reported. For
modeling the consequences of vapor cloud explosions, a "yield factor" is applied to account for the
unexploded vapor in the cloud.
6.1.4 Type of Ignition
The energy supplied to the gas/air mixture per unit time and per unit volume by an ignition source
is extremely important in determining whether ignition immediately results in a detonation or in a
deflagration. Much greater energy is required for the initiation of a detonation. According to the TNO
Yellow Book, the minimum ignition energy required for detonation is unknown for most gases, but the
first test results for some hydrocarbon gases indicate that this value can vary by orders of magnitude.
6.1.5 Delayed Ignition Effects
Time between failure of chemical containment and ignition of a vapor cloud can vary by incident
(Exhibit 3) depending on the concentration of released vapors and the proximity of those vapors to an
ignition source. Wiekema (1984) noted that "...an ignition within about one minute after the beginning of
the release will enhance the possibility of an explosion." According to the M & M data base, the longest
reported time before ignition (35 minutes) occurred in a facility in Feyzin, France. Seven mile-per-hour
winds are thought to have been instrumental in directing the butane vapors away from facility ignition
sources. The cloud was eventually ignited by an automobile on a road nearby the plant boundary (M & M
Data Base 1990).
-------
-16-
Exhibit 3
TIME BEFORE IGNITION
Facility
Pampa, TX
Norco, LA
Pasadena, TX
New Castle, DE
Texas City, TX
Baton Rouge, LA
Mount Belvieu, TX
Pernis, Netherlands
Abqaiq, Saudi Arabia
Mexico City, Mexico
Feyzin, France
Time Between Failure and
Ignition
10 seconds
30 seconds
1 minute
1.5-2 minutes
2 minutes
2 minutes
4-5 minutes
6-8 minutes
7 minutes
10 minutes
35 minutes (windy)
Source: M & M Data Base, 1990
Wiekema (1984) reported that over 60 percent of vapor cloud explosions are reported to result
from ignition within 100 meters of the release site. However, some vapor clouds may not explode until an
ignition source is found in a nearby community, possibly increasing the chance of off-site damage. Exhibit
4 shows a number of incidents in which vapor clouds ignited at some distance from the point of release.
Delayed ignition might allow more flammable vapor be released, and the vapor cloud might travel
further off-site, possibly increasing the potential for off-site damage. However, the vapor cloud will also
disperse with time. According to Kletz (1977), if ignition is delayed, the quantity within the explosive
range will first increase and then decrease as the vapor disperses:
-------
-17-
Exhibit 4
DISTANCE TO IGNITION
Facility
Billings, MT
Goi, Japan
Denver, CO
Feyzin, France
Texas City, TX
Abqaiq, Saudi Arabia
Rio De Janeiro, Brazil
Chemical
butane
propylene/hexane
propane
butane
hydrocarbons
gas
LPG
Distance to' Ignition
30m
45m
90m
*
90m
195m
450m
beyond facility boundary
Source: M & M Data base, 1990
6.2 Vapor Cloud Fires
Vapor cloud fires, also referred to as flash fires, do not generate extreme overpressures similar to
vapor cloud explosions. Instead, the associated hazard is a rapidly moving flame front. The size of the
vapor cloud is determined by the extent to which the released material disperses in the atmosphere before
ignition. If a vapor cloud or plume contacts an ignition source at a point at which its concentration is
within the range of its upper and lower flammable limits, a wall of flame may flash back towards the
source of the gas or vapor (FEMA, DOT, EPA 1989). Following ignition of an .over-rich cloud of fuel, a
diffusion-type flame rolls back across the cloud of fuel as air is mixed further with the over-rich cloud until
the entire cloud is burned (Lewis 1980). Individuals engulfed in such a vapor cloud that then ignites have
a high probability of being killed. However, there were no reported fatalities beyond the actual vapor
cloud due to radiation effects in a sample of vapor cloud fires examined by Wiekema (1984). He
concludes that since the area outside of the cloud exposed to intense thermal radiation is small, no
fatalities due to thermal radiation can be expected outside the area engulfed in flame.
One incident of a vapor cloud fire occurred at an Austin, Texas, pump station on February 22,
1973. Failure of a ten-inch pipeline carrying natural gas liquids (NGL) at 525 psig pressure caused the
release and vaporization of 278,880 gallons of NGL onto a road adjacent to the pump station and into
ditches adjacent to the road. Shortly after the release, two cars entered the vapor zone and stalled; the
occupants of these vehicles immediately fled safely out of the vapor zone. Soon after, a van carrying six
adults and two children stalled. The passengers exited the vehicle and the driver tried to restart the van's
engine. A spark ignited the fuel-air mixture, immediately killing four of the van's occupants; two
occupants later died of injuries sustained, while the other two were severely burned. All three vehicles
were destroyed. A 2,400 foot long section of pasture land along the road was charred black (Eisenberg,
1975).
6.3 BLEVEs and Resulting Fireballs
Boiling Liquid Expanding Vapor Explosions (BLEVEs) occur when sealed tanks of liquid or
gaseous hazardous materials are exposed to fire, which may cause excessive pressures within the tank
-------
-18-
combined with weakening of tank walls. The sudden failure of the vessel and rapid vaporization and
expansion of its contents is termed a BLEVE. BLEVEs also generally result in ignition of the vapor cloud
when the substance is flammable, and a large rising fireball may form, the size of which will vary with the
accident conditions and the type and amount of hazardous material present. Although the fireball is
generally of short duration, the intense thermal radiation generated can cause severe and possibly fatal
burns to exposed people over relatively considerable distances in a matter of seconds (FEMA, DOT, EPA
1989). Overpressures and container fragment projectiles also may be generated by BLEVEs but are of less
concern than the thermal radiation (Nazario 1988). See Section 6.6 for a discussion of projectile hazards.
For all major fuel release incidents studied which led to BLEVEs, the incident was accompanied by a
major fire incident (Lewis 1980).
Lewis (1980) describes BLEVEs for fuels as follows:
When a pressure vessel containing a liquefied flammable gas or a flammable liquid is heated
by an external fire which heats the metal wall at the vapor space level, the vessel will rupture
into a number of large pieces which will rocket considerable distances as fuel inside the
pieces burns. This is accompanied by a large fireball and some explosive pressure effects
produced from the liquid rapidly expanding during the propagation of fractures as the vessel
ruptures. These pressure effects are usually minor compared with the fires started by heat
radiation.
The Center for Chemical Process Safety Guidelines for Chemical Process Quantitative Risk
Analysis (CCPS 1989) describe a fireball as follows:
Fireball: The atmospheric burning of a fuel-air cloud in which the energy is mostly emitted
in the form of radiant heat. The inner core of the fuel release consists of almost pure fuel
whereas the outer layer in which ignition first occurs is a flammable fuel-air mixture. As
buoyancy forces of the hot gases begin to dominate, the burning cloud rises and becomes
more spherical in shape.
Nazario (1988) suggests a guideline that if 10% of a chemical vaporizes when it is released to the
atmosphere from a vessel, the chemical has a high potential to BLEVE. Pentane and lighter hydrocarbons
generally satisfy this vaporization criterion when contained in vessels at their typical storage temperatures
and pressures and released to the atmosphere. Results produced by the models for chemicals heavier than
pentane have therefore not been considered further in this assessment; however, it should be noted that it
is not impossible for less volatile chemicals, when subjected to elevated temperature and pressure
conditions, to result in BLEVEs.
BLEVEs are not always the principal cause of disaster but can occur secondarily when, for example,
flames from an ignited vapor cloud impinge upon containers with flammables/explosive contents. Exhibit 5
contains a sampling of chemical incidents involving BLEVEs that have occurred in the last thirty years. In
all of these incidents, other types of fires and explosions were associated with the BLEVEs.
Several catastrophic accidents have been classified as BLEVEs. One of the most notable was at a
Mexico City liquefied petroleum gas (LPG) site where BLEVEs occurred involving millions of pounds of
flammable gas and numerous tanks. Hundreds of residents in homes adjacent to the plant were killed.
-------
-19-
Exhibit 5
INCIDENTS INVOLVING BLEVEs IN THE M & M DATABASE
Incident Location,
Date
Type of Facility
Cause/Description
Damage
Pasadena, TX
(10/23/89)
Petrochemical
Plant
Vapor cloud of ethylene and isobutane formed and ignited approximately one
minute after release. Heat from fireball thought to have caused BLEVEs of
neighboring pressure tanks.
$500 to
S750M
Romeoville,
Illinois (07/23/84)
Refinery
Explosion of released propane resulted in fire engulfing the unsaturated gas plant,
FCC, and alkylation units. Fire burned out of control for about 30 minutes before a
BLEVE occurred in a process vessel in the alkylation unit One piece of debris
thrown 500 feet cut pipelines and finally struck a tank in the water treatment unit.
Another piece of debris was thrown 600 feet where it caused another major fire.
S143.5M
Mexico City,
Mexico (11A9/84)
Terminal
Release of LPG during transfer caused formation of LPG vapor cloud. Structures
at the facility included six spheres and 48 bullets for a total storage capacity of more
than 4.2M gallons of LPG. Within five minutes of ignition of the LPG vapor cloud,
the first of a series of BLEVEs produced a fireball approximately 1200 feet in
diameter. Heat and projectiles from the BLEVE released fuel from other tanks.
Four of the six spheres (each with 420,000 gallon capacity and about half-full at time
of transfer line rupture) and 44 of the bullets BLEVE'd or were ruptured by
missiles. The two largest spheres, full to their 630,000 gallon capacity were saved
from BLEVE by water from 100 rail tank cars. Contents of these two spheres then
burned under controlled conditions.
S22.5M
Priola, Italy
(05A9/85)
Petrochemical
Plant
Fire from ignition of escaped hydrocarbon caused explosion of a tall vertical
propane tank, skyrocketing the top section 1,500 feet and missing a gas holder by 30
feet. The deluge water-spray system apparently was not effective under the given
fire situation.
S72.8M
Texas City, TX
(05/30/78)
Refinery
Within 20 minutes of ignition of vapor cloud, five 1,000 barrel horizontal bullets,
four 1,000 barrel vertical bullets, and two additional 5,000 barrel spheres failed from
missile damage or BLEVEs. Pieces of these tanks fell into other units, starting
more fires.
S93M
Houston, TX
(10A9/71)
Derailment of twenty tank cars led to puncture of a vinyl chloride tank. Released
fuel ignited. A vinyl chloride tank car containing less than 145 tons of vinyl chloride
BLEVEd 45 minutes later, killing one and injuring 50 people. Fireball reported as
305 meters in diameter and scattered debris up to 91 meters away.
Crescent City, IL
(06/21/70)
Ten rail cars with carrying capacity of 33,000 U.S. gallons of propane apiece
derailed. Propane from a punctured tank car was released and ignited causing six of
the tank cars to rupture and BLEVE. Sixty-six people were injured and business
section of town destroyed.
$3M
New Jersey
Turnpike, 1 mile
south of exit 8
(09/21/72)
Friction sparks ignited fuel leaking from tractor-trailer fuel tank and spread to
propylene leaking from cargo tank, engulfing a substantial portion of cargo tank in
flame. Twenty to 25 minutes later, the cargo tank exploded, burning or otherwise
injuring 28 people, one of whom was 600 feet away from the explosion. Two people
trapped in an automobile wedged between the tractor-trailer and guard-rail were
killed. A 27-foot long piece of the trailer tank was thrown 1307 feet northeast and
400 feet east of the explosion center.
Sources:
Houston, TX and Crescent City, IL incidents from Lewis, David J. "Unconfined Vapour-Cloud Explosions-Historical Perspective and Predictive
Method Based on Incident Records." 1980. Prog. Energy Comb. Sci., Vol. 6, pp. 151-165. Pergamon Press, Ltd.
New Jersey Turnpike incident from Eisenberg et at., Vulnerability Model: A Simulation for Assessing Damage Resultine from Marine Spills.
June 1975. Prepared for Department of Transportation, U.S. Coast Guard Office of Research and Development.
Other incidents from M & M Protection Consultants, 1990.
-------
-20-
6.4 Pool Fires
A liquid pool fire is a fire involving a quantity of liquid fuel spilled on the surface of the land or
water. An ignition source must start the pool fire. Primary hazards of liquid pool fires to people and
property involve exposure to thermal radiation and/or toxic or corrosive products of combustion (FEMA,
DOT, EPA 1989).
One incident involving a pool fire occurred at Hearne, Texas, on May 14,1972. Rupture of an
eight-inch crude oil pipeline caused the release of oil onto the surface of a nearby river. Oil on the
surface of the water reportedly travelled through culverts under a highway and railroad and collected on a
stock pond located 1,800 feet from the point of the rupture. After ignition of the oil vapors by an
unidentified source, a fire began along the stream and burned back through the culverts and to the stock
pool. Described as "1,800 feet long and several hundred feet high", the fire stopped all highway and
railroad traffic. The fire was fueled by 7,913 barrels of released crude oil (Eisenberg, 1975).
6.5 Jet Fires
I
Transportation and storage tanks or pipelines containing compressed or liquefied gases may
discharge gases at a high speed if somehow punctured or ruptured during an accident. The gas discharging
or venting from the hole will form a gas jet that "blows" into the atmosphere in the direction the hole is
facing, all the while entraining and mixing with air. When the gas is flammable and it encounters an
ignition source, a flame jet of considerable length may form. For chemicals that are liquid under ambient
conditions, no gas jet will form; liquid or vapor might leak out through a puncture or break, but will not
blow out.
6.6 Projectiles
Projectiles resulting from BLEVEs and VCEs present potential hazards in addition to the hazards
of heat radiation and blast overpressures. Projectiles can kill people, pierce chemical tanks/reactors, sever
chemical and electrical pipelines, destroy fire fighting equipment, and even start fires far from the
explosion. Frequently, projectiles compound an already catastrophic situation.
According to AIChE (1987), vessels that fail under pressure (e.g., in BLEVEs) generally produce a
few large fragments. Possible fragmentations include cylindrical straight walls breaking into one or two
metal sheets; the head and/or bottom breaking away from the cylinder section; and bolted and threaded
inserts (thermocouples, pressure gauges) failing and generating missiles. According to Pineau et al. (1991),
the mean weight of fragments from accidental explosions is of the order of 30 to 100 kg (70 to 220
pounds). Confined detonations, on the other hand, may form a relatively large number of primary
fragments or shrapnel (AIChE 1987).
The M & M data base includes numerous examples of incidents with projectiles (Exhibit 6). In
Grangemouth, United Kingdom, a 3-ton piece of a pressure separator was thrown 1,000 meters from an
explosion. In addition, an incident in Feyzin, France, caused pieces of steel weighing up to 100 tons to be
thrown 1.2 kilometers (three-quarters of a mile). One of these fragments landed on a pipeline and cut 40
lines. These incidents illustrate the danger of projectiles resulting from explosions involving flammable
substances and the need to consider them in consequence analysis. However, consideration of potential
damage from projectiles is largely site-specific and therefore is not incorporated into general consequence
models such as WHAZAN. Analysis of projectile effects must include explosion pressure, rate of pressure
rise, vessel shape, size and material of construction, as well as velocity, penetration, and range of
fragments.
-------
-21-
Exhibit 6
PROJECTILES FROM EXPLOSION INCIDENTS
Facility
Geismar, LA
Priola, Italy
Texas City, TX
Grangemouth, UK
Romeoville, JL
Mexico City, Mexico
Feyzin, France
Type of Explosion
VCE
fire and tank
explosion
heat decomposition
overpressure in
separator
VCE
BLEVE and VCE
VCE, fire, and tank
explosion
Projectile Weight
reactor head
top section of a
distillation column
0.4 tons
3 tons
20 tons
20 tons
100 tons
Projectile Distance
425m
450m
*900m
1,000m
1,070 m
1,190m
1,200 m
Source: M & M Data Base, 1990
Two examples, shown in Exhibits 1 and 8, illustrate the complexity in estimating projectile effects.
Exhibit 7, taken from AIChE course materials (1987), analyzes the overpressurization of an ethoxylation
reactor in which a flange and a vessel fragment separate during the explosion. Note that the equation for
calculating projectile distance assumes no air drag, and is thus conservative if the projectile does not have
an aerodynamic shape and glides. Exhibit 8 includes an example from the Yellow Book (TNO 1980), in
which a propane vessel explodes under ductile fracture. Exhibit 8 also presents the ethoxylation
overpressurization example from Exhibit 7, calculated using the Yellow Book method.
Because the AIChE and Yellow Book methods are based on different assumptions and have
different data requirements, a number of assumptions had to be made to carry out the Yellow Book
calculations for the ethoxylation overpressurization example, as noted in Exhibit 8. For Scenario 1 (failure
of a 30 pound flange), a distance of 34 meters was calculated using the AIChE method, while the Yellow
Book method gave 70 meters, approximately twice the distance calculated by the AIChE method. For
Scenario 2, where a 150 pound vessel fragment is removed from the reactor head by pressure energy, the
AIChE method gave a distance of 1.25 kilometers, while the Yellow Book method gave a much shorter
distance of 40 meters. These calculations are presented only as examples; the calculation methods have
not been evaluated or compared in detail.
-------
-22-
Exhibit 7
SAMPLE CALCULATIONS FOR PROJECTILES (AIChE)
EXAMPLE-EXPLOSION FRAGMENTS
A 3000 gallon reactor used for ethoxylation operates at 70°C and 85 psig. The reactor contains 100 ft3 of vapor
space above the reaction mass. The reactor fails catastrophically at 400 psig (414.7 psi), 4 times the design pressure.
Calculate the explosion effects from two ejected-material scenarios.
Scenario 1. A 30 pound blind flange on a 4 inch nozzle falls. Estimate fragment velocity (feet per second) and
distance (feet).
Initial Velocity
v0 = 2.05 [ (PD3) / W]w
(where: v0= velocity (ft/s), P = internal pressure (psig), D = diameter of failed nozzle (in), W = weight of projectile
(Ib)) t
v0 - 2.05 [ (400)(4)3) / 30)0-5
v0 = 60 ft/s (18 m/s)
Distance
R = [v02sin2(a)]/g
(where: R = radial distance (ft), a = angle of projectile path (degrees), g = gravitational constant (ft/s2))
R = [602 $^2(45°) ]/32.2
R - 112 ft (34 m)
Scenario 2. A 150 pound vessel fragment with surface area of 10 ft2 (total vapor space surface area = 120 ft2) is
removed from the reactor head by pressure energy. Estimate fragment velocity and distance.
Total Pressure Energy is calculated using the equation:
E = 1.26 (V) (P,/P0)
(where: V = volume (ft3), P0 = standard pressure (psi), P, = pressure of compressed gas (psi), P2 = final pressure of
expanded gas (psi), T0 - standard temperature (K), Tl - temperature of compressed gas (K), R - universal gas
constant (1.987 cal/g-mol-K)
E = 1.26 (100) (414.7/14.7) (273) (1.987) ln(414.7/14.7) = 6,439,561 cal
The AJChE method assumes 20% of total energy (KE, kinetic energy) is used to shatter the vessel. The kinetic
energy fraction expended on the vessel fragment is needed (the fragment is 10 ft2 of a total 120 ft2 surface area).
KE = (6,439461) (3.086 ft-lb/cal) = 19,872,485 ft-lb
KE = (19,872,485) (0.2) (10A20) = 331,208 ft-lb
KE = (%mv02)/g
Solve for v0:
v0-[2(KE)(g)/m]a5
v, ° [ 2(331.208)(32.2) / ISO \°-s - 377 ft/s (115mA)
Distance (using same equation as in Scenario 1)
R = 3772 sin2(45°) / 32.2 = 4114 ft (1.25 km)
Therefore, results for Scenario 1 estimate that the 30 pound flange would be ejected at an initial velocity of 60
ft/sec, and travel a distance of 112 feet (34 meters). Scenario 2 results indicate the 10 square foot vessel fragment
would initially separate at 377 ft/sec, and travel 4,114 feet (1,250 meters).
Source: AIChE Course Materials
-------
-23-
Exhibit 8
SAMPLE CALCULATIONS FOR PROJECTILES (YELLOW BOOK)
EXAMPLE A-WORKED PROJECTILE
Propane is contained in a carbon steel vessel, having 0.01S m thickness, 4 m diameter, 8 m length (100 m3
volume), and weighing IS metric tons. A 0.15 m diameter nozzle is connected to the vessel, at a height of 10 m. The
vessel has an operating pressure of 2 MPa, and bursts under ductile fracture at 5 MPa. Operating temperature is
60°C. There is 10 m3 of liquid and 90 m3 vapor in the vessel. Other relevant properties are listed below:
vapor liquid
t = 1.25 Pl = 600 kg/m3
explosion pressure = 8 MPa Cp/h,, = 0.005
boiling point = -40°C
The nozzle separates bom the vessel during the explosion. Find its initial velocity, and the distance the nozzle travels.
The following equation is used for calculating initial velocity of a fragment:
uto = [(2/m) x (F/(l + i3"3) x (Ap/(Y-l) x (»/4) x d2hja5
(where: uf 0 = initial velocity (m/sec), m = mass of vessel (kg), F = yield factor for fragment energy (0.6 for ductile
fracture, 0.2 for brittle), e = fraction of strain of the vessel material on fracture (0.38 for carbon steels), Ap = change
in pressure (Pa), y = ratio of specific heats at constant pressure and constant volume (gases), d = diameter of
cylindrical vessel (m), h = height of vapor space above any liquid present (m))
ufo = [(2A5000) x (0.6/(l+0383a25)'3) x (3xlO*/(1.25-l) x (r/4) x 425]as
uto = 260 m/s
Maximum distance the projectile travels is found by iteratively solving this equation for r
r^ - 0.23 x h'w x uto [exp (-C. x ( V»r) * P, 01
(where: rmaj = maximum distance from center of vessel (m), h'= ejection height (m), uto= initial velocity (m/sec),
Q,= coefficient of resistance, Af= area of fragment (m2), mf= mass of fragment projectile (kg), pa = density of air
(kg/m3),.r = distance from center of vessel (m))
By substituting the equations for nozzle area for Af, and nozzle volume multiplied by vessel material density for mr, the
ratio Af/njf can be simplified to 8-55xlO~Vd.
r^ = 0.23 x 1005 x 260 [exp (-1.11 x (S-SSxlfl-4 / 0.15) x 1.29 r)]
By iteratively choosing values of r, a value of r is found for which the equation is true:
rnm = r = 90 m
The nozzle travels about 90 m from the vessel.
Source: Yellow Book, Bureau for Industrial Safety (TNO), The Netherlands
-------
-24-
Exhibit 8 (continued)
SAMPLE CALCULATIONS FOR PROJECTILES (YELLOW BOOK)
EXAMPLE B-CALCULATING SCENARIOS 1 & 2 FROM EXHIBIT 7 BY YELLOW BOOK METHOD
Using this Yellow Book Method, we will now calculate Scenarios 1 and 2 from Exhibit 7 to provide a
comparison of the results obtained by each method. Since the Yellow Book method is dependent on the size and
shape of the rupturing container, we will first calculate the reactor size and total weight for a reactor with a
capacity of 3,000 gallons or approximately 1136 m3 of space. Assuming the reactor height = 1.5(diatneter of the
reactor), the appropriate reactor dimensions would be a height (h')= 4.02m and diameter = 2.68m. |ince there is
100 ft3 of vapor space above the liquid in the vessel, the height of vapor above the liquid is h = 1.004 m. Assume
the reactor is constructed of carbon steel and that the fracture is ductile F-0.6. The weight of the entire reactor
can now be calculated by multiplying the total surface area of the reactor by 150 lbs/10 fr (based on the vessel
fragment described in Scenario Z) The weight of the entire reactor is then 6373 Ibs or 2891 kgs.
The ratio of specific heats (y) for ethylene oxide is 1.21 @ 293 K. The P^,,, given was 414.7 psi or (4 x
design capacity), therefore P,.^,^, is assumed to be (10 x design capacity) per Yellow Book for worst case
scenario or 1036.8 psi, and AP (Pe^touon-Pbum) is 4-29 * 10* Pa- Assume Q, = 1.11 for both Scenarios 1 and 2.
Since the Yellow Book calculation of uto is based on the vessel from which the projectile originated and is
independent of the actual projectile which is ejected, the initial velocity calculation of uLo is identical for the flange
in Scenario 1 and the reactor fragment in Scenario 2. The initial speed of both the flange and the reactor fragment
can be calculated:
uto - [(2/2891) x (O^l+O-SS3'1-21*-3)) x (4.29xl06/(1.21-l)) x (»/4) x 2.68^1.004]^
"to = 197 m/s.
The Yellow Book method is specific to the size and shape of the ejected object only when the deceleration
term is added in the calculation of r,^ The maximum distance the projectile travels is found by iteratively solving
this equation for r
r^ = 0.23 x h'as x U(;o [exp (-Q, x (Aptaf) x p3 r)].
Scenario 1 Assuming the flange has a diameter of 8" or 0.2032m (2 x the nozzle diameter), the surface area of
the flange (Af) can be calculated as 0.0324 m. The mass of the flange (mr) was given in Scenario 1 as 30 Ibs or
13.61 kg. Finally, assuming Cw~l.ll, r^caa be calculated:
rmal = 0.23 x 4.02a5 x 197 (exp (-1.11 x (0.0324 / 13.61) x 1.29 r)].
By iteratively choosing values of r, a value of r is found for which the equation is true:
Maximum distance travelled by the flange r,,,,, = r = 70 m.
Scenario 2 The area of the vessel fragment (Af)is given in Scenario 2 as 10 ft2 or 0.929 m2. The mass (mf) of the
vessel fragment was also given in Scenario 2 as 150 Ibs or 68.04 kg. Finally, assuming Cw=l.ll, r^can be
calculated:
TO,, - 0.23 x 4.02*5 x 197 [exp (-1.11 x (0.929 / 68.04) x 1.29 r)J.
By iteratively choosing values of r, a value of r is found for which the equation is true:
Maximum distance travelled by the reactor fragment r,,,^ = r = 40 m.
-------
-25-
7.0 Accident Factors and Model Inputs for Consequence Analysis
Accident scenario factors (e.g., concentration, confinement, release pressure) could have an effect
the severity of the consequences of an accident involving a flammable chemical. A few factors such as
quantity released, temperature/pressure conditions, and atmospheric stability are common inputs to most
consequence models. However, most models are limited in the number of factors that can be input into
the consequence analysis. It is impossible to list all of these factors, especially those that are site-specific
(e.g., characteristics of terrain and local micrometeorological effects such as slight wind shifts). The effect
of these factors on the consequences of flammable chemical accidents is difficult to measure, however, this
section will qualitatively evaluate this effect. Later in this section, the factors and modeling inputs that
can be incorporated into the models will be discussed. *
7.1 Model Inputs
The WHAZAN model (discussed in Section 5.1 and in Appendix B, Section B.I) was applied to
flammable liquids and gases to determine the distance that might be affected for various types of accidents,
using varying quantities released and various hazard criteria levels. WHAZAN is not sufficiently complex
to incorporate certain accident factors such as flashback and projectiles for vapor cloud explosions.
However, several WHAZAN model inputs that will help determine accident consequences include the
following:
• Specific physical/chemical properties of the chemical involved (e.g., volatility and heat
of combustion);
• The quantity of the chemical involved;
• The type of release (instantaneous or prolonged);
• The initial conditions (temperature and pressure) of the chemical just prior to release;
• Meteorology (particularly when modeling scenarios that involve dispersion); and
• The surface roughness over which the release passes.
These factors and their relationship to consequence modeling are discussed below.
7.1.1 Chemicals and Quantities Chosen for Modeling
A variety of chemicals were modeled for selected quantities. Ten common hydrocarbons, including
aikanes, alkenes, and aromatics, of widely varying volatility (boiling points from -103°C to 137°C) and
flammability (as indicated by flash point) were chosen for consequence modeling as representative
examples of a range of flammable chemicals that may be found in commerce in high volumes. Volatility
and flammability are important characteristics, in determining accident consequences. The selected
chemicals included ethylene, propane, propylene, n-butane, pentane, hexane, heptane, toluene, p-xylene,
and gasoline. Physical/chemical properties of these chemicals are listed in Exhibit E-2, Appendix E.
Ethylene, propane, propylene, and n-butane are gases under ambient conditions; the other chemicals are
liquids. WHAZAN modeling was carried out for releases of 1,000, 2,500, 5,000, 7,500, 10,000, 25,000,
50,000, 100,000 and 180,000 pounds of each of the ten hydrocarbons to gain insight into the impact of the
size of an accidental release on potential consequences. Results of modeling are presented in Section 8.
Modeling of 100,000 pound releases of each of the ten hydrocarbons was carried out using the other
methods discussed in Section 5 to compare the results from other models (PC-based calculation methods
-------
-26-
and others) with WHAZAN. The methods for all the models used are discussed in Appendix B. Results
of modeling by all methods were found to be in reasonable agreement. Eight other organic chemicals,
including ethylene oxide, ethyl chloride, acetaldehyde, acetone, trichloroethylene, acetic acid, benzaldehyde,
and dimethyl sulfoxide were also modeled for 100,000 pound releases using the WHAZAN model for
comparison with the hydrocarbons. Some of these chemicals (e.g., trichloroethylene and dimethyl
sulfoxide) are not considered very flammable; they were included for comparison purposes. Results of
modeling 100,000 pound releases of the hydrocarbons by all methods and WHAZAN results for the
additional chemicals are presented in Exhibit D-l, Appendix D.
7.1.2 Instantaneous and Prolonged Releases
. t
Instantaneous releases are those that take place over a period of time ranging from a few seconds
to several minutes and then essentially stop. The result of such a release typically is a liquid pool, a puff
of vapor or gas, or distinct cloud, but can also involve an explosion. A sudden catastrophic vessel failure
or the failure of a large piping connection can be considered examples of instantaneous releases.
Instantaneous releases were considered likely to have potentially more severe consequences than prolonged
releases (see below) and were assumed as the type of release in the comparative modeling using all
methods.
Prolonged or continuous releases take place over longer periods of time and, when atmospheric
dispersion is involved, typically produce plumes stretched out over longer distances. Failure of a small
piping connection or of a piping gasket are examples of potential prolonged releases. For modeling
purposes, prolonged releases were assumed to emanate from 1.5 inch diameter holes in containment
vessels. This simulates the shearing off of a small instrument connection on a vessel or a gasket leak in a
larger piece of piping. Prolonged releases were modeled using WHAZAN for comparison with
instantaneous releases; the consequences are, in general, smaller.
7.1.3 Initial Conditions Prior to Release
Flammable chemicals are handled and/or stored at a wide variety of temperatures and pressures at
industrial facilities. Conditions can range from extremely severe processing conditions (high temperature
and pressure) to relatively low severity storage conditions. Very volatile chemicals may be stored as liquids
under pressure and/or at low temperatures. Less volatile chemicals can be stored at ambient temperatures
and pressures. Accident consequence modeling is dependent on initial release conditions. Because it is
not uncommon for storage vessels to be sited closer to plant fencelines and hence neighboring
communities than processing equipment, and because storage vessels generally contain much larger
volumes of material than processing equipment, storage temperatures and pressures that may be found in
industry practice were assumed as initial release conditions for modeling purposes. Temperatures and
pressures used in modeling are listed by chemical in Exhibit E-l, Appendix E.
7.1.4 Meteorology
The primary meteorological conditions used for modeling by all methods were 3.0 meters per
second wind speed and D atmospheric stability, representing average or neutral atmospheric conditions
rather than worst case situations. More extreme conditions (for dispersion) of 1.5 meters per second wind
speed and F atmospheric stability were also used for WHAZAN modeling to show the effect of varying
meteorological conditions on accident consequences. As explained in the Green Book (EPA, FEMA,
DOT 1987), these conditions correspond to stable air and low wind speed, conditions under which
dispersion results for distances to various hazard criteria levels are generally larger than those calculated
under conditions usually considered typical.
-------
-27-
7.1.5 Surface Roughness
Surface roughness influences the dispersion and gravity spreading of dense gas clouds. It is a
measure of the irregularity of the terrain over which a cloud passes. Irregularities.include mountains,
trees, flat rural areas, and city buildings. The size of the surface feature relates to the size of a surface
roughness factor (e.g., large irregularities are assigned large factors). The surface roughness factor of 0.1
meters is the WHAZAN model default value that is consistent with long grass conditions.
8.0 Hazard Criteria
The extent and severity of the impacts of an accident depends on the type of consequence that
results. Accidents involving flammable substances can result in thermal radiation from fires or blast
overpressures from explosions, or both. The consequences for various types of accidents have been
summarized below:
Accident * Consequence Units
Vapor Cloud Explosion Overpressure psi
BLEVE Thermal radiation, duration of exposure kW/m2, seconds
Pool Fire Thermal radiation, duration of exposure kW/m2, seconds
Jet Fire Thermal radiation, duration of exposure kW/m2, seconds
Vapor Cloud Fire Ignition of fire LFL
where psi = pounds per square inch, kW/m2 = kilowatts per square meter, and LFL = lower flammable
limit; these are discussed further below.
For modeling purposes, several hazard criteria levels of overpressure, thermal radiation, or
concentration associated with each type of accident consequence were used for modeling. The effects of
the hazard criteria and the distances resulting from modeling can then be used by EPA in regulatory
determinations. For a given accident, the WHAZAN model was used to calculate the distance from the
accident site to the chosen hazard criteria level. The hazard criteria levels that were used in modeling may
have effects that range from minor injury to humans through various levels of property damage to
potentially lethal effects on humans (see Appendix C). These are discussed below by type of accident.
8.1 Hazard Criteria for Vapor Cloud Explosions
The consequence of greatest concern for vapor cloud explosions is the blast wave. Blast waves
from accidental explosions can cause damage to people and property by subjecting them to transient
crushing pressures and winds. Relatively simple concepts have been used to correlate blast wave
properties with damage. Blast wave damage is primarily a function of either the peak overpressure, the
impulse, or some combination of these factors. Peak overpressure is most commonly used. For modeling
vapor cloud explosions, overpressures of 0.3, 0.5, 1.0, 3.0, and 10.0 psi were considered. At an
overpressure of 0.3 psi, 10 percent of window glass may be broken (Lees 1980). An overpressure level of
0.5 psi can cause windows to shatter, with some frame damages (Brasie and Simpson 1968). Overpressure
of 1.0 psi can cause partial demolition of houses and is the threshold for skin laceration from flying glass
(Lees 1980). An overpressure level of 3.0 psi can cause steel frame buildings to become distorted and
pulled away from their foundations (Lees 1980); this overpressure can also hurl a person to the ground
(Brasie and Simpson 1968). An overpressure of 10 psi may totally .destroy buildings (Lees 1980) and is the
threshold for lung hemorrhage (U.S. DOT 1988). Pineau et al. (1991) suggest overpressure levels of 170
mbar (2.5 psi) as the threshold for significant lethality and 50 mbar (0.7 psi) as the upper limit for
reversible effects on humans, based on a review of accidents in France. Appendix C, Section C.1.1,
-------
-28-
presents additional information on damage and injury caused by various overpressure levels from
explosions.
8.2 Hazard Criteria for Vapor Cloud Fires
The minimum concentration of a vapor or gas in air that will ignite and propagate flame is known
as its lower flammable limit (LFL) concentration or lower explosive limit (LEL) concentration. The words
flammable and explosive are used interchangeably such that LFL values equal LEL values; the
concentration of a fuel that will burn in air also can be expected to explode under the appropriate
conditions. Similarly, the maximum concentration of a gas or vapor in air that will ignite and propagate
flame is known as the upper flammable limit (UFL) or upper explosive limit (UEL) of*the fuel (FEMA,
DOT, EPA 1989).
For vapor cloud fires, dispersion modeling was carried out using concentrations equal to 50
percent of the LFL, the LFL, and twice the LFL. Levels below the LFL are used to account for variability
within the cloud from its edge to its center and inaccuracies in dispersion modeling. This is based on
guidance in the WHAZAN manual (WHAZAN 1988) which recommends the following:
When calculating safe distances from flammable releases, these difficulties [uncertainties
in concentration] indicate that the analyst should consider concentrations somewhat below
the LFL. A concentration of LFL/2 or LFL/4 is normally used.
8.3 Hazard Criteria for BLEVEs, Pool Fires, and Jet Fires
The consequence of greatest concern resulting from BLEVEs, pool fires, and jet fires is heat
radiation. In the case of BLEVEs, injuries and fatalities might also result from overpressures and
container fragment projectiles, but thermal radiation is considered to be the consequence of greatest
concern and is the consequence addressed in the modeling. Human injury or fatality from heat radiation
from fires is likely to occur as a result of direct exposure to a fire rather than as a result of property
damage caused by a fire. The extent of the injury caused by heat radiation depends both on the heat level
and the time of exposure. The duration of the fire and the length of time it might take an exposed person
to escape or take shelter from the fire would be important considerations.
The WHAZAN model produces results for duration of the fire and distance from the fire for
three heat radiation levels for BLEVEs, pool fires, and jet fires. For BLEVEs and jet fires, duration and
distance results are produced for heat radiation levels of 4.0,12.5, and 37.5 kW/m2 According to the
WHAZAN manual, a heat radiation level of 4.0 kW/m2 would lead to a 1 percent probability of fatality
after an exposure time of 150 seconds. For 12.5 kW/m2 the probability of fatality is 1 percent after a 30
second exposure and 50 percent after an exposure time of 80 seconds. For a heat radiation level of 37.5
kW/m2 the probability of fatality is 1 percent for an 8-second exposure, 50 percent for a 20-second
exposure, and 99 percent for a 50-second exposure (WHAZAN 1988). For pool fires, WHAZAN gives
results for 1.6, 4.0, and 12.5 kW/m2. A heat radiation level of 1.6 kW/m2 would lead to a 1 percent
probability of fatality after 500 seconds (WHAZAN 1988).
Pineau et al. suggest heat radiation thresholds of 5.2 kW/m2 for severe casulaties and lethality and
2.9 kW/m2 as the upper limit for reversible radiation effects for a fire lasting 60 seconds or more. These
thresholds are intended to apply when escape from the heat effects is not possible. Pineau's threshold
levels seem consistent with the heat radiation information presented in the WHAZAN manual. Appendix
C, Section C.3.1, presents additional information on thermal radiation.
-------
-29-
9.0 Modeling Results
WHAZAN modeling was carried out for vapor cloud explosions, vapor cloud fires, BLEVES, pool
fires, and jet fires of the hydrocarbons using the assumptions discussed above. For vapor cloud explosions,
data from actual events described in the M & M data base were compared with results predicted by
WHAZAN. All WHAZAN results are presented in the following sections by consequence type. Resulting
distances for selected quantities and several hazard criteria levels are provided. Results of modeling by
other methods and additional WHAZAN results are presented in Exhibit D-l, Appendix D. The inputs
used for the WHAZAN model are presented in detail in Appendix E; Appendix E also lists inputs for the
ARCHIE model. As noted earlier, results from all the methods are in reasonable agreement.
»
9.1 Vapor Cloud Explosion Results
9.1.1 Modeling Results for Four Hydrocarbons
WHAZAN Mo'del Used. To obtain the WHAZAN modeling results presented here, the linked
models were used only to determine the fraction of pressurized or refrigerated gas instantly flashed to
vapor (the model calculates fraction flashed only for substances that are gases under ambient conditions).
The fraction flashed was used to estimate the flammable mass in the cloud; this quantity was used to run
the stand-alone model for vapor cloud explosion, rather than the linked models. The linked models were
not used for vapor cloud explosions because it was found that these models did not produce reasonable
results for this type of consequence; i.e., the quantity apparently assumed to be in the cloud seemed
unreasonably high (much higher than the quantity that would be immediately flashed into vapor).
Quantity in Cloud. For modeling purposes, the quantity in the cloud was assumed to be the
fraction flashed. According to Kletz (1977), the adiabatic flash gives an estimate of the minimum amount
of material present in the cloud. Some material is likely to be released into the cloud as aerosol, but
WHAZAN does not provide a method for estimating aerosolization. The entire quantity in the cloud was
assumed to be within the explosive limits. In an actual incident, the quantity in the cloud might be greater
than the fraction flashed, and the quantity within the explosive limits might be smaller than the total
quantity in the cloud because of dispersion. The most conservative assumption would be to assume the
entire quantity is in the cloud; such an occurrence seems very unlikely, however.
Yield Factor. The explosive yield factor was assumed to be 0.11, the default value for WHAZAN
(i.e., 11 percent of the quantity in the cloud was assumed to participate in the explosion). This is intended
to be a conservative assumption. The actual explosive yield in a vapor cloud explosion may vary greatly,
depending on the substance involved, storage and release conditions, and other factors. According to Lees
(1980), yield factors are usually between one and 10 percent; however, cases of much higher efficiencies
have been reported (e.g., 25 to 30 percent in one incident) (Lees 1980).
Overpressure Levels. WHAZAN results for vapor cloud explosions are presented as distances to
four characteristic damage levels; overpressures related to these damage levels are not presented but have
been estimated for this report. Affected distances calculated by WHAZAN appear to be based on the
scaling law of distances, which states that distance is proportional to the cube root of quantity of explosive
material. For TNT, empirical graphs of scaled distance versus overpressure have been derived. The
WHAZAN distances are based on "limit values" related to damage levels. It was assumed, based on the
scaling law, that the WHAZAN distances would be proportional to the scaled distance values for TNT as
presented in Lees (1980), the source cited in the WHAZAN manual. Using this relationship and the
descriptions of the damage levels in the WHAZAN manual, the Yellow Book, and Lees, the overpressure
related to each limit value was estimated.
-------
-30-
Results. Exhibit 9 presents WHAZAN modeling results for vapor cloud explosions of four
hydrocarbon gases for quantities of 1,000 to 180,000 pounds. Ethylene and propane are assumed to be
liquefied by refrigeration; propylene and n-butane are assumed to be liquefied under pressure. Vapor
cloud explosion results could not be obtained for liquids under ambient conditions. The exhibit shows the
fraction flashed, obtained from the linked models, and the quantity assumed to be in the cloud, calculated
from the total quantity released and the fraction flashed. Distances calculated for several overpressure
levels are presented. Although the modeling shows distances of approximately 100 meters for release
quantities as small as about 5,000 pounds at overpressures of 1.0 psi and lower, the probability of a vapor
cloud explosion occurring is relatively low for vapor clouds containing quantities smaller than 10,000
pounds (API 1990, Prugh 1987), as noted earlier. Note also that the quantity in the cloud is considerably
smaller than the quantity released; i.e., a release of 10,000 pounds of flammable material would probably
result in a cloud containing much less than 10,000 pounds (Exhibit 9 shows 150 to 3,300 pounds in the
cloud for a release of 10,000 pounds). Distances for ethylene and propane, assumed to be liquefied by
refrigeration, are considerably shorter than for propylene and n-butane, assumed to be liquefied under
pressure, probably because at reduced temperatures less of the ethylene and propane are in the vapor
phase to contribute to the amount flashed upon release. For propylene and n-butane, modeling results
indicate distances greater than 100 meters for an overpressure of 1.0 psi for all the quantities modeled.
For an overpressure of 1.0 psi, distances for propane range from 49 meters for a 1,000 pound release to
276 meters for a 180,000 pound release, while for propylene the distances are from 139 meters for 1,000
pounds to 787 meters for 180,000 pounds.
Exhibit 9 shows results for ethylene and propane under low-temperature storage conditions.
Ethylene and propane were also modeled assuming initial release conditions of higher temperatures and
pressures, similar to the conditions assumed for propylene and n-butane; under these conditions, distances
were similar to those found for propylene and n-butane, indicating that gases liquefied by refrigeration may
pose less potential risk of vapor cloud explosion than gases liquefied under pressure.
Exhibit 10 shows WHAZAN modeling results as a graph of quantity released versus distance for
vapor cloud explosions, based on 1.0 psi overpressure. It is assumed that graphs of quantity versus
distance for other overpressure levels would show similar increases in distance with increasing quantity
released. Quantities released displayed on the graph range from 1,000 to 50,000 pounds. The curves
shown in Exhibit 10 illustrate the fact that distance is proportional to the cube root of the quantity of
explosive, according to the scaling law of distances, which is the basis for WHAZAN modeling of vapor
cloud explosions. Thus, at a specific hazard criteria level, the smaller the quantity released, the much
smaller the distance for experiencing the impact.
Exhibit C-6, in Section C.1.2 of Appendix C, shows WHAZAN modeling results for releases of
100,000 pounds of ethylene and propylene. Distances to different overpressure levels are displayed
graphically. These results indicate that the choice of overpressure level has a significant effect on
determining the affected distance, particularly at low overpressures. For example, for propylene the
affected distance decreases from 1,121 meters to 647 meters if an overpressure of 1.0 psi is used rather
than 0.5 psi. Similar decreases occur for the other chemicals. The graph of distance versus overpressure
becomes relatively flat between overpressures of 3.0 and 10.0.
-------
-31-
Exhibit 9
DISTANCES FOR VAPOR CLOUD EXPLOSIONS OF FOUR LIQUEFIED GASES
FOR INSTANTANEOUS RELEASES OF 1,000 TO 180,000 POUNDS
DETERMINED USING FRACTION FLASHED, WHAZAN STAND-ALONE MODEL
Chemical .
(Initial
Conditions)
Ethylene
(liquefied by
refrigeration
175 K, 1.3
Bars.)
Propylene
(liquefied
under
pressure
293 K, 13.3
Bars)
Propane
(liquefied by
refrigeration
232 K, 1.3
Bars)
n-Butane
(liquefied
under
pressure
293 K, 6.3
Bars)
Quantity
Released
(pounds)
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
Fraction
Flashed
0.023
0.329
0.015
0.113
Quantity
in Cloud
(pounds)
23
58
115
173
230
575
1,150
2,300
4,140
329
823
1,645
2,468
3,290
8,225
16,450
32,900
59,220
15
38
75
113
150
375
750
1,500
2,700
113
283
565
848
1,130
2,825
5,650
11,300
20,340
»
Distance in Meters to the Following Overpressures
0.3 psi
151
205
259
2%
326
443
558
703
855
372
494
622
712
801
1,090
1,370
1,730
2,100
130
177
223
255
280
381
479
604
735
275
345
434
497
546
713
936
1,180
1,430
0.5 psi
98
134
168
193
212
288
362
457
556
242
321
404
463
521
706
890
1,121
1,364
85
115
145
166
182
247
312
393
478
179
224
282
323
355
463
607
766
933
1.0 psi
57
77
97
111
122
166
209
263
321
139
185
233
267
300
408
513
647
787
49
66
83
96
105
143
180
227
276
103
129
163
186
205
267
350
442
538
3.0 psi
23
31
39
44
49
66
84
. 105
128
56
74
93
107
120
163
206
259
315
20
26
33
38
42
57
72
91
110
41
52
65
75
82
107
140
177
215
10.0 psi
11
15
19
22
24
33
42
53
64
28
37
47
53
60
' 82
103
129
157
10
13
17
19
21
29
36
45
55
21
26
33
37
41
53
70
88
108
-------
600
500
400
~ 300
•
o
I
° 200
100
-*-
Exhibit 10
QUANTITY RELEASED VERSUS DISTANCE
FOR VAPOR CLOUD EXPLOSIONS, INSTANTANEOUS RELEASES
WI1AZAN Modeling for 1,000-50,000 pounds, 1.0 psi Overpressure
Ethylene*
**
-S Propylene
-B-— Propane*
-------
-33-
Issues. Several factors used in WHAZAN to model the consequences of a vapor cloud explosion
may vary widely in an actual event; therefore, the results of an actual incident may differ from the results
presented here. The modeling could have been carried out using more or less conservative assumptions,
and larger or smaller results would have been obtained. For this modeling, a conservative yield factor of
11% was used; generally, yields are likely to be lower, but higher yields have also been reported. Use of
the fraction flashed as the quantity in the cloud is not a conservative assumption, because additional
material is likely to be carried into the cloud. Assuming the entire cloud is within the explosive range is
conservative, however.
9.1.2 Modeling Using Other Meteorological Conditions
*
WHAZAN modeling was also carried out using meteorological conditions of 1.5 meters per
second wind speed and F atmospheric stability. This was done to show the impacts of varying
meteorological conditions on accident consequences. Results for vapor cloud explosions were identical to
those obtained using 3.0 meters per second wind speed and D atmospheric stability, indicating that the
vapor cloud explosion results using WHAZAN are not dependent on meteorological assumptions made.
These results are presented in Exhibit D-l, Appendix D.
9.1.3 Modeling Prolonged Releases
WHAZAN modeling for a prolonged release was also undertaken to compare the results of
assuming an instantaneous versus a prolonged release. However, this attempt, which used moderate
meteorological conditions, produced no vapor cloud explosion results. It was not possible to determine
why the model produced no vapor cloud explosion data for prolonged releases. ARCHIE modeling was
then tried for modeling prolonged releases (assuming a 1.5-inch hole) of 1,000 to 10,000 pounds of
propylene. In the case of a liquid release, the results for both instantaneous and prolonged releases were
essentially identical; therefore, these results are not presented here. If a prolonged gas release was
assumed, the model calculated the same distances for each different quantity released. Releases of other
gases would probably provide similar results. These results are not logically consistent and the particular
reason(s) for the inconsistencies are not known. However, this may be a function of the limitations of the
ARCHIE model at small release quantities. As discussed earlier, vapor cloud explosions are more likely if
ignition occurs within one minute of a release (Wiekema 1984); under these circumstances, prolonged
release models are not appropriate.
9.2 Comparison of Vapor Cloud Explosion Data and WHAZAN Results
To verify the general assumptions used and the results obtained from the vapor explosion
algorithms contained in WHAZAN, actual vapor cloud incidents were modeled by the WHAZAN Stand
Alone Model. The M & M data base provided the information on vapor cloud incidents (e.g., chemical,
quantity released) which formed the basis of the model inputs. Similar to the previous analysis, the
WHAZAN default value for explosion yield of 11 percent was used. Five vapor cloud explosion incidents
involving four different chemicals were studied. Exhibit 11 compares the overpressure and distance results
from the WHAZAN model with the overpressures and distances calculated or measured in the actual
explosions. Overpressures for the incidents modeled are reported in the M & M data base or were
estimated based on the descriptions of damage. The greatest distance at which glass was broken was
assumed to occur at 0.3 psi. In the modeling, the use of fraction flashed was not necessary because the
accident description already provided an estimate of the quantity in the cloud rather than the quantity
released. The five vapor cloud explosions are described further following Exhibit 11.
-------
-34-
Exhibit 11
COMPARISON OF EXPLOSION DATA AND WHAZAN RESULTS
Incident
Case#l:
(Pasadena)
Case #2:
(Norco)
Case #3:
(East St. Louis)
Case #4:
(Port Hudson)
Case #5:
(Flixborough, UK)
Chemical
(Initial
Conditions)
Ethylene
700 psi
Propane
- 270 psi
and 130° F
Propylene
Liquified
Propane
942 psig
Cyclohexane
155" C
Quantity
Released
into Cloud
(pounds)
85,000
20,000
107,000
132,000
60,000
WHAZAN RESULTS
(Distance in Meters '
to the Following Overpressures)
0.3 psi
2340
1430
2500
2690
2020
1.0 psi
878
537
938
1010
757
3.0 psi
351
215
375
403
303
10.0 psi
175
107
188
202
152
EXPLOSION
DATA
(Based on
Reported
Damage)
7 psi at .
100 meters'
as high as 10 psi
near epicenter
(within 100
meters)
0.15 - 0.4 psi
at 1000 meters
0.3 psi at
8000 meters
10 psi at 135
meters
3 psi at 335 to
535 meters;
0.3 psi at 2400
meters
Based on estimated TNT equivalence of 10 tons
Notes: 0.3 psi is assumed to be overpressure level for glass breakage.
Explosion yield factor of 11% is assumed for above cases.
Atmospheric conditions are not needed as input to WHAZAN Stand Alone Model.
Case #1 Comparison In one accident in Pasadena, Texas, 85,000 pounds of a mixture containing
primarily ethylene was released through a valve at 700 psi. The vapor cloud ignited after
approximately 1 minute. The blast was equivalent to the detonation of 10 tons of TNT, or
approximately 7 psi overpressure at 100 meters. A run of the WHAZAN model with similar
inputs gave a range of overpressures and distances (e.g., 10 psi overpressure at 175 meters and 3
psi overpressure at 351 meters). The explosion data was just outside this predicted range. Due
the exponential relationship between the overpressure and distance, this difference between the
WHAZAN results and the explosion data may not be significant
Case #2 Comparison In a refinery vapor cloud explosion in Norco, Louisiana, 20,000 pounds of
propane were released through a failed pipe at 270 psi and 130° F. A vapor cloud formed during
the 30 seconds between failure and ignition. The resulting explosion of propane caused
overpressures as high as 10 psi within 100 meters. The 10 psi overpressure was reported in the M
-------
-35-
& M data base (1990) based on analysis of facility equipment in the blast path. Using the
WHAZAN model, an overpressure of 10 psi at 107 meters was predicted.
Case #3 Comparison Rail cars collided in East St. Louis, Illinois releasing 107,000 pounds of
propylene. An elongated cloud covering about 5 acres was formed before'an ignition source
ignited the cloud after about a 5 minute delay. The resulting vapor cloud explosion caused an
overpressure of between 0.15 and 0.4 psi at 1 kilometer from the explosion center. At this
distance, a school suffered extensive interior damage. A run of the WHAZAN model with similar
inputs gave 0.3 psi at 2.5 kilometers.
Case #4 Comparison In Port Hudson, Missouri, a pipeline ruptured and released ^32,000 pounds
of liquefied propane at 942 psig into a cloud. A vapor cloud, covering about 10 acres, was ignited
after 24 minutes. The resulting explosion broke windows 8 kilometers away, overpressure for
window breakage was assumed to be 0.3 psi. A run of the WHAZAN model with similar inputs
gave 0.3 psi at only 2.7 kilometers.
Case #5 Comparison In a petrochemical plant in Flixborough, United Kingdom, a massive failure
of a 20-inch diameter bypass assembly released approximately 60,000 pounds of primarily
cyclohexane at 155° C. The huge vapor cloud, measuring 120 by 150 meters, was ignited. The
resulting explosion and fire destroyed much of the plant. At 120 meters from the center of the
explosion, the walls of a brick building collapsed. The vapor cloud explosion at Flixborough has
been studied extensively and therefore has much damage data (Sadee 1977). The overpressure was
between approximately 10 psi at 135 meters, 3 psi at 335 to 535 meters, and approximately 0.3 psi
(between 10 and 20 percent of windows were damaged) at 2,400 meters away. A run of the
WHAZAN model with similar inputs gave comparable results with 10 psi at 152 meters, 3 psi at
about 300 meters, and 0.3 psi at about 2000 meters.
Exhibit 11 shows that the WHAZAN model results for cases #1, #2, #3, and #5 generally
correlate with the consequences of a vapor cloud explosions. The correlations in cases #3 and #5 are
stronger than cases #1 and #2 because overpressures at distances further from the explosion center are
more easy to predict and measure than distances closer to the explosion center. However, in three of
these cases, WHAZAN results were consistently more conservative than the explosion data (at same
overpressure, predicted distances using WHAZAN are greater than distances measured in actual
explosions).
The generally conservative WHAZAN results in cases #1, #2, and #3 may be due to the
potentially conservative assumption of 11 percent explosive yield. An explosive yield factor of 2 percent
was mentioned as a reasonable estimate by several members of EPA's Science Advisory Board on
flammable chemicals. The 2 percent explosion yield factor produces results which more closely simulate
the explosion data. However, the 11 percent factor used to calculate vapor cloud explosion results in
Section 9.1 is acceptable because for this analysis, conservative estimates of affected distances is desired.
In case #4, the WHAZAN results seem to underestimate the consequences significantly.
However, this vapor cloud explosion was not a typical unconfined aerial explosion ignited by a spark or
flame. Instead, pan of the cloud entered a warehouse and was ignited by a spark. The warehouse
explosion became the powerful initiator for the explosion of the unconfined vapor cloud surrounding the
warehouse. According to Lewis (1980), this type of explosion may be described as a "quasi-detonation;"
the damage is much nearer to that observed with a condensed phase explosion than the damage given by a
typical unconfined aerial explosion. Because the WHAZAN model does not model this type of intensified
explosion, the consequence modeling may have underestimated the effects.
-------
-36-
9.3 Vapor Cloud Fire Results
WHAZAN Model Used. Vapor cloud fires were modeled using WHAZAN by running the linked
WHAZAN models to obtain the pool evaporation rate. The pool evaporation rate was used as the release
rate for the stand-alone buoyant plume dispersion model. The WHAZAN buoyant plume dispersion
model was used to estimate the greatest downwind distances at which a flame will move through a
flammable vapor cloud. The greatest distance affected by a vapor cloud fire is assumed to be the distance
at which the cloud concentration is equal to the lower flammable limit. Any person located within the
limited area covered by the flammable part of the vapor cloud would be engulfed in the ensuing fire and
would likely die. For a given downwind distance, the total area (as a function of downwind and crosswind
distance) potentially affected by a vapor cloud fire is likely to be much smaller than the'area that might be
affected by a vapor cloud explosion.
Note that the buoyant plume dispersion model uses the release rate (assumed to be the pool
evaporation rate), not. the fraction flashed. The results shown here do not include the quantity of gas
immediately flashed into vapor on release, but only the quantity that evaporates from a pool after the
initial flash. For propylene and n-butane, gases assumed to be liquefied under pressure, the fraction
flashed is appreciable (33 percent for propylene, 11 percent for n-butane), indicating the calculated
distances for these materials may be understated, because the material that flashes could also be involved
in the fire.
Results. Exhibit 12 displays the results of dispersion modeling using WHAZAN for vapor clouds
of flammable liquids and gases. The data presented include the downwind and crosswind distances at
which the concentration in the cloud is equal to 50 percent of the LFL, the LFL, and twice the LFL. The
downwind distance is the maximum distance at which the specified concentration is reached; the crosswind
distance is the width of the cloud of flammable vapor at that point. The modeled crosswind distances are
much smaller than the downwind distances, roughly 6 percent, for the chemicals and conditions analyzed.
The largest modeled downwind distance results (approximately 1,800 meters) were for a release of
180,000 pounds of ethylene; a 1,000 pound release of p-xylene showed the smallest downwind distance
(about 20 meters) results. Ethylene also provided the largest downwind distance and p-xylene the smallest
at the other hazard criteria levels analyzed for vapor cloud fires.
Exhibit 13 presents WHAZAN vapor cloud fire results as a graph of quantity released versus
distance for instantaneous releases, assuming moderate meteorological conditions and modeling to a
concentration equal to 50 percent of the LFL. Distances are clearly not linearly related to the quantity
released; i.e., increasing the quantity released by a factor of ten increases the distance by a factor of
approximately two and one-half for the chemicals modeled.
Effect of Varying Concentration. Exhibit C-7, in Section C.2 of Appendix C, is a graphical
presentation of results based on 50 percent of the LFL, the LFL, and twice the LFL for vapor cloud fires
for two hydrocarbons. This analysis was done to determine the sensitivity of the model results to changes
in the flammable limit level. Using the LFL rather than 50 percent of the LFL for 100,000 pounds of
ethylene lowers the affected distance, as would be expected, but by much less than a factor of two. The
effect of changing from the LFL to twice the LFL is even smaller. The data presented in Exhibit 12 show
there is relatively little difference in distance based on varying the hazard criteria levels for the other
chemicals analyzed.
-------
-37-
Exhibit 12
DISTANCES FOR VAPOR CLOUD FIRES FROM EVAPORATING POOLS
FROM INSTANTANEOUS RELEASES, MODERATE METEOROLOGY*,
DETERMINED USING WIIAZAN
Distance in Meters to Three Concentrations
Chemical(Initia)
Conditions)
Ethylene
(liquefied by refrigeration
175 K, 1.3 Bars)
•
Propylene
(liquefied under pressure
293 K, 13.3 Bars)
Propane
(liquefied by refrigeration
232 K, 1.3 Bars)
Quantity
Released
(pounds)
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
Evaporation
Rate
(Ibs/sec)
60
115
190
254
313
611
1,012
1,678
2,580
20
42
68
90
110
214
353
587
900
24
49
79
108
132
260
430
712
1,091
1/2 LFL
Downwind
Distance
(meters)
203
294
391
462
522
768
1,030
1,390
1,810
106
158
207
242
270
394
527
705
. 905
112
162
214
254
285
417
557
747
959
Crosswind
Distance
(meters)
13
18
24
28
31
45
58
77
97
7
10
13
15
17
24
31
41
52
7
10
13
16
18
25
33
43
54
LFL
Downwind
Distance
(meters)
138
199
264
311
351
514
688
925
1,190
74
107
141
164
184
266
354
473
605
77
110
145
172
193
282
375
500
640
Crosswind
Distance
(meters)
9
13
16
19
21
31
40
53
67
5
7
9
10
12
17
22
28
36
5
7
9
11 •»
12
17
23
30
38
2 LFL
Downwind
Distance
(meters)
95
135
179
211
237
346
462
618
793
60
87
112
131
147
212
281
375
479
55
77
99
117
131
191
253
337
430
Crosswind
Distance
(meters)
6
9
11
13
15
21
28
36
46
4
6
7
9
9
13
17
23
29
4
5
7
8
9
12
16
21
26
Wind speed 3.0 meters per second, atmospheric stability class D.
-------
-38-
Exhibit 12 (continued)
DISTANCES FOR VAPOR CLOUD FIRES FROM EVAPORATING POOLS
FROM INSTANTANEOUS RELEASES, MODERATE METEOROLOGY*,
DETERMINED USING WHAZAN
Distance in Meters to Three Concentrations
Chemical(lnitia)
Conditions)
n-Butane
(liquefied under pressure
293 K, 6.3 Bars)
Pentane
(293 K, 2.74 Bars)
Gasoline
(293 K, 1.013 Bars)
Quantity
Released
(pounds)
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
Evaporation
Rate
(Ibs/sec)
11
24
40
53
64
126
209
346
534
7
15
24
33
40
77
128
212
326
7
13
22
29
35
71
117
194
298
1/2 LFL
Downwind
Distance
(meters)
73
103
135
159
178
259
345
460
588
53
81
101
120
133
193
256
341
435
55
78
100
115
132
, 191
254
338
431
Crosswind
Distance
(meters)
5
7
9
10
11
16
21
28
35
4
5
7
8
9
12
16
21
26
4
5
7
8
9
12
16
21
26
LFL
Downwind
Distance
(meters)
49
73
93
108
120
175
233
309
396
38
57
72
84
92
131
173
230
294
39
55
71
81
89
130
172
228
290
Crosswind
Distance
(meters)
3
5
6
7
8
11
15
19
24
3
4
5
6
6
9
11
14
18
3
4
5
5"
6
8
11
14
18
2 LFL
Downwind
Distance
(meters)
35
52
66
76
84
119
158
209
267
27
41
51
59
65
90
118
156
199
28
39
50
57
63
89
117
155
196
Crosswind
Distance
(meters)
2
3
4
5
6
8
10
13
17
2
3
3
4
4
6
8
10
13
2
3
3
4
4
6
8
10
12
Wind speed 3.0 mete.rs per second, atmospheric stability class D.
-------
-39-
Exhibit 12 (continued)
DISTANCES FOR VAPOR CLOUD FIRES FROM EVAPORATING POOLS
FROM INSTANTANEOUS RELEASES, MODERATE METEOROLOGY*,
DETERMINED USING WHAZAN
Distance in Meters to Three Concentrations
Chemical (Initial
Conditions)
Hexane
(293 K, 1.013 Bars)
Heptane
(293 K, 1.013 Bars)
Toluene
(293 K, 1.013 Bars)
Quantity
Released
(pounds)
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
Evaporation
Rate
(Ibs/sec)
4
9
13
18
22
44
73
121
185
2
4
9
11
13
26
42
71
108
2
4
7
9
9
20
33
53
82
1/2 LFL
Downwind
Distance
(meters)
41
58
71
82
91
133
176
233
297
29
41
58
65
71
99
130
173
219
31
43
53
61
64
89
117
154
196
Crosswind
Distance
(meters)
3
4
5
5
6
9
11
15
18
2
3
4
4
5
7
9
11
14
2
3
4-
4
4
6
8
10
12
LFL
Downwind
Distance
(meters)
29
41
50
58
65
91
119
158
201
21
29
41
46
50
71
89
118
150
22
31
37
43
43
64
83
105
134
Crosswind
Distance
(meters)
2
3
3
4
4
6
8
10
13
1
2
3
3
3
5
6
8
10
1
2
2
3
3 *
4
6
7
9
2 LFL
Downwind
Distance
(meters)
21
29
36
41
46
65
83
108
136
15
21
29
33
36
50
63
82
101
16
22
27
31
31
46
59
74
92
Crosswind
Distance
(meters)
1
2
2
3
3
4
6
7
9
1
1
2
2
2
3
4
5
7
1
1
2
2
2
3
4
5
6
Wind speed 3.0 meters per second, atmospheric stability class D.
-------
-40-
Exhibit 12 (continued)
DISTANCES FOR VAPOR CLOUD FIRES FROM EVAPORATING POOLS
FROM INSTANTANEOUS RELEASES, MODERATE METEOROLOGY*,
DETERMINED USING WHAZAN
Distance in Meters to Three Concentrations
Chemical(lnitial
Conditions)
p-Xylene
(293 K, 1.013 Bars)
Quantity
Released
(pounds)
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
Evaporation
Rate
(Ibs/sec)
1
2
4
4
7
1.1
20
31
49
1/2 LFL
Downwind
Distance
(meters)
21
30
42
42
48
67
86
111
141
Crosswind
Distance
(meters)
1
2
3
3
3
4
6
7
9
LFL
Downwind
Distance
(meters)
15
21
30
30
36
47
63
78
97
Crosswind
Distance
(meters)
1
1
2
2
2
3
4
5
7
2 LFL
Downwind
Distance
(meters)
11
15
21
21
26
33
44
55
69
Crosswind
Distance
(meters)
1
1
1
I
2
2
3
4
5
Wind speed 3.0 meters per second, atmospheric stability class D.
-------
1200
Exhibit 13
QUANTITY RELEASED VERSUS DISTANCE
FOR VAPOR CLOUD FIRES, INSTANTANEOUS
RELEASES, MODERATE METEOROLOGY
WHAZAN Modelling for 1,000-50,000 Pounds, 50% LFL
1000
800
•
•X
E
u
600
400
20O
J.
1-
e. . ,_
A
A
*
Ethylene
**
Propylene
Propane*
n-Butane* *
Pentane
Gasoline
Hexane
20 30 40
Quantity Released (thousands of pounds)
-------
-42-
Modeling Using Other Meteoroloeicai Conditions. WHAZAN modeling was also carried out for
the ten chemicals shown in Exhibit 11 using additional meteorological conditions of wind speed of 1.5
meters per second and F atmospheric stability as inputs, for comparison with the results obtained using the
average meteorological conditions. The WHAZAN results for vapor cloud fires appear to be strongly
dependent on the meteorological assumptions made, with distances determined using worst case
meteorology being much greater than those resulting from modeling using moderate meteorology.
Distances for a concentration equal to 50 percent of the LFL under worst case meteorological conditions
extend from 1,790 to 6,350 meters for gases, compared to 460 to 1,390 meters under moderate conditions.
These results are presented in Exhibit D-l, Appendix D.
Prolonged and Instantaneous Releases. Exhibit D-l, Appendix D, also includes Results of
WHAZAN modeling for prolonged releases of ten chemicals, where release was assumed to take place
from 1.5 inch diameter holes in 12 foot diameter tanks. Distances to the 50 percent of the LFL level for
vapor cloud fires are much smaller (18 to 79 meters) than for instantaneous releases, indicating that using
instantaneous release conditions is, in general, more conservative than using prolonged release conditions.
Issues. Results of dispersion modeling can vary greatly depending on assumptions used.
Meteorological conditions can have a very large effect. In addition, the results presented in Exhibit 12 are
based on the assumption that the cloud of flammable vapor ignites when it has reached the maximum
distance to the specified concentration; in an actual incident, ignition might occur at any time following
the release, or the cloud could disperse without igniting.
9.4 BLEVE Results
WHAZAN Model Used. BLEVE results were obtained using the linked WHAZAN models; the
stand-alone BLEVE model was found to give the same results. The WHAZAN default combustion
efficiency factor of 0.263 was assumed.
Results. Exhibit 14 presents WHAZAN modeling results for BLEVEs of gases (ethylene,
propylene, propane, and n-butane) and for pentane, a volatile liquid assumed to be stored under pressure,
for release quantities of 1,000 pounds to 180,000 pounds. Distances for three levels of heat radiation and
the duration of the BLEVEs are shown. These data are all important components in analyzing the
consequences of BLEVEs, which are a function of exposure to heat radiation levels for the duration of the
resulting fireball. As noted in Section 6.1.3, pentane and lighter hydrocarbons have characteristics (i.e., 10
percent of the chemical vaporizes when it is released to the atmosphere from a vessel) that give them a
high potential to be involved in BLEVEs (Nazario 1988). While heavier hydrocarbons may be subject to
BLEVEs under some conditions, the likelihood of occurrence of a BLEVE is much smaller for such
chemicals; therefore, WHAZAN BLEVE results for hydrocarbons heavier than pentane are not included
in Exhibit 14.
The most volatile of the chemicals analyzed, ethylene, shows the greatest distances, and the least
volatile, pentane, the smallest; however, distances do not vary greatly from chemical to chemical. For
release quantities of less than 10,000 pounds, at a heat radiation level of 12.5 kW/m2, distances for
ethylene vary from 82 meters for 1,000 pounds to 179 meters for 10,000 pounds. For pentane, the range is
56 meters for 1,000 pounds to 120 meters for 10,000 pounds.
Draft September 29, 1993 ***
-------
-43-
Exhibit 14
DISTANCES FOR BLEVES
FOR INSTANTANEOUS RELEASES OF 1,000 TO 180,000 POUNDS
DETERMINED USING WHAZAN
Distance in Meters to Three Heat Radiation Levels
Chemical
(Initial
Conditions)
Ethylene
(liquefied by
refrigeration
175 K, 1.3 Bars)
Propylene
(liquefied under
pressure
293 K, 13.3 Bars)
Propane
(liquefied by
refrigeration
232 K, 1.3 Bars)
Quantity
Released
(pounds)
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
Distance in Meters to the Following Heat
Radiation Levels
4.0 kW/m2
146
198
250
287
316
430
543
685
835
108
147
185
212
234
318
402
508
618
104
142
179
206
227
308
389
492
599
12.5 kW/m2
82
112
141
162
179
243
307
388
472
61
83
105
120
132
180
227
287
350
59
80
102
116
128
174
220
278
339
37.5 kW/m2
47
65
82
94
103
140
177
224.
273
35
48
60
69
76
104
131
166
202
34
46
59
67
74
101
127
161
196
BLEVE
•Duration
(seconds)
8
11
14
16
17
23
29
36
44
8
11
14
16
17
23
29
36
44
8
11
14
16
17
23
29
36
44
-------
-44-
Exhibit 14 (continued)
DISTANCES FOR BLEVES
FOR INSTANTANEOUS RELEASES OF 1,000 TO 180,000 POUNDS
DETERMINED USING WHAZAN
Chemical
(Initial
Conditions)
n-Butane
(liquefied under
pressure
293 K, 6.3 Bars)
Pentane
(293 K,
2.74 Bars)
Quantity
Released
(pounds)
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
Distance in Meters to the Following Heat
Radiation Levels
4.0 kW/m2
101
137
173
199
219
298
376
475
579
101
137
173
199
219
298
376
475
579
12.5 kW/m2
57
78
98
112
124
169
213
269
328
57
78
98
112
124
169
213
269
328
37.5 kW/m2
. *
33
45
57
65
72
97
123
155
189
33
45
57
65
72
97
123
155
189
BLEVE
Duration
(seconds)
8
11
14
16
17
23
29
36
44
8
11
14
16
17
23
29
36
' 44
-------
-45-
The duration of a BLEVE as estimated by WHAZAN appears to depend only on the quantity
released; it does not vary by chemical. For release quantities of 10,000 pounds or less, WHAZAN shows
the BLEVE duration to range from 8 seconds for 1,000 pounds to 17 seconds for 10,000 pounds. As
noted earlier, the WHAZAN manual (WHAZAN 1988) suggests that at a heat radiation level of 12.5
kW/m2, there would be a 1 percent probability of fatality for a 30 second exposure. Therefore, for
BLEVEs lasting less than 30 seconds, as is the case for BLEVEs involving quantities less than 10,000
pounds, a higher heat radiation level might be of greater interest (e.g., 37.5 kW/m2 rather than 12.5
kW/m2), as exposure could not last longer than 20 seconds. For a heat radiation level of 37.5 kW/m2,
distances for ethylene are 39 meters for a 1,000 pound release and 103 meters for a 10,000 pound release.
Exhibit 15 shows WHAZAN results, presented graphically, for BLEVEs of 1,000 to* 50,000 pounds
of hydrocarbons. The graph of released quantity versus distance indicates that distance is proportional to
the cube root of quantity, as was the case for vapor cloud explosions.
Effect of Varying Heat Radiation Levels. Exhibit C-10, in Section C.3.2 of Appendix C, presents
BLEVE results, in graphical form, from WHAZAN modeling of two hydrocarbons for three heat radiation
levels, to test the model results for sensitivity to varying hazard criteria levels. For ethylene and propylene,
using 37.5 kW/m2 rather than 12.5 kW/m2 appears to have a relatively small effect on results; using 4
kW/m2 rather than 12.5 kW/m2 appears to have a greater effect.
Modeling Using Other Meteorological Conditions. WHAZAN modeling was also carried out for
BLEVEs using worst case meteorological conditions. Varying the meteorological conditions appeared to
have no effect; results obtained were identical to the results obtained using moderate meteorological
conditions, indicating that meteorological conditions would be expected to have little impact on the
consequences of BLEVEs. Results are presented in Exhibit D-l, Appendix D.
Prolonged and Instantaneous Releases. WHAZAN modeling assuming a prolonged release for a
BLEVE gave results identical to those assuming an instantaneous release, as shown in Exhibit D-l,
Appendix D. As BLEVEs result from sudden vessel failure, they would always be essentially
instantaneous; therefore, it is not clear that any distinction can be made between instantaneous versus
prolonged release input in modeling BLEVEs.
Issues. As discussed above, the duration of the fireball from a BLEVE, as well as the heat
intensity, is an important consideration for estimating potential consequences of a BLEVE.
In addition to the heat radiation effects, BLEVEs can also result in projectiles. Section 6.6
discusses the characteristics and hazards of projectiles from BLEVEs. The distances affected by projectiles
may be larger than the distances affected by heat radiation; for example, calculations presented by AIChE
(1987) (see Exhibit 7) show a 150 pound vessel fragment thrown more than 1,000 meters from a 3,000
gallon reactor. This distance is greater than distances calculated to any heat radiation level by WHAZAN
modeling. The distances traveled and the potential effects of projectiles, however, are more localized,
explosion/site specific, and unpredictable.
9.5 Pool Fire Results
WHAZAN Model Used. The WHAZAN linked models were used to obtain pool fire results.
For substances that are gases under ambient conditions, the linked models calculate the fraction flashed
and treat the remaining liquid as a circular pool that spreads to maximum size. For liquids, the entire
quantity released is assumed to be in the pool. The WHAZAN default value for combustion efficiency of
0.35 was assumed.
-------
350
300
250
<«^
M
| 200
»*
S ISO
100
50
EihlMt 15
QUANTITY RELEASED VERSUS DISTANCE
FOR BLEVES, INSTANTANEOUS RELEASES
WHAZAN Modelling for 1,000-50,000 Pounds, 115 kW/m2
-* Ethylene*
-O— Propylene'
-Q— Propane*
-A-— n-Butane
.**
1
1
1
10
20 30 40
Quantity Released (thousands off pounds)
50
-------
-47-
Results. Exhibit 16 presents WHAZAN modeling results for pool fires resulting from releases of
1,000 to 180,000 pounds of ten hydrocarbons. Distances are given for three levels of heat radiation. The
initial quantity in the pool and the duration of the fire are also provided.
For all the chemicals shown in Exhibit 16, distances for releases smaller than 10,000 pounds are
less than 100 meters. The distance results for pool fires do not vary greatly by chemical, possibly because
all of the chemicals shown here are hydrocarbons, and all have similar heats of combustion. (WHAZAN
produced somewhat smaller distances for a given quantity of several organic chemicals with differing heats
of combustion; see Exhibit D-l, Appendix D.) Ethylene and propane, the two gases assumed liquefied by
refrigeration, show greater distances than propylene and n-butane, gases assumed to be liquefied under
pressure. Larger quantities of the pressurized gases flash into vapor on release than for theirefrigerated
gases; therefore, the quantity in the pool would be larger for the refrigerated gases, and the greater
distances would be expected.
The pool fire duration data range from about 20 seconds for 1,000 pounds of propylene to
approximately 2 minutes for 180,000 pounds of p-xylene. After that time the material in the pool is
consumed, extinguishing the pool fire. The data for a specific quantity and hazard criteria level are fairly
consistent from chemical to chemical.
The WHAZAN pool fire results are presented graphically in Exhibit 17 for releases of 1,000 to
50,000 pounds. The graph of quantity released versus distance indicates that distance is proportional to
the cube root of the quantity released, as was the case for vapor cloud explosions and BLEVEs.
Effect of Varying Heat Radiation Levels. Exhibit C-ll, in Section C.3.2 of Appendix C, presents
WHAZAN modeling results, in graphical form, for ethylene and propylene, showing the effect of using
different heat radiation levels. As was the case for BLEVEs, using a different heat radiation level appears
to have a greater effect on distance at lower levels than at higher levels. The graph indicates that basing
results on 10 kW/m2, as in ARCHIE modeling (see Exhibit D-l, Appendix D), rather than 12.5 kW/m2,
probably does not have a great effect on the results.
Modeling Using Other Meteorological Conditions. WHAZAN modeling of pool fires for ten
chemicals using worst case meteorology gave results identical to those obtained using moderate
meteorology, indicating that meteorological conditions probably do not have much effect on pool fire
consequences, These results are presented in Exhibit D-l, Appendix D.
Prolonged and Instantaneous Releases. WHAZAN pool fire results for prolonged releases
(assuming release from a hole with a diameter of 1.5 inches) of 100,000 pounds of ten chemicals are
presented in Exhibit D-l, Appendix D; results for instantaneous releases are also included for comparison.
Distances for prolonged releases are much smaller (9 to 23 meters compared to 152 to 182 meters) than
for instantaneous releases, again indicating that the input of instantaneous release type is more
conservative than assuming a prolonged release.
-------
-48-
Exhibit 16
DISTANCES FOR POOL FIRES
FOR INSTANTANEOUS RELEASES OF 1,000 TO 180,000 POUNDS
DETERMINED USING WHAZAN
Distance in Meters to Three Heat Radiation Levels
Chemical
(Initial
Conditions)
Ethylene
(liquefied by
refrigeration
175 K, 1.3 Bars)
Propylene
(liquefied under
pressure
293 K, 133 Bars)
Propane
(liquefied by
refrigeration
232 K, 13 Bars)
n-Butane
(liquefied under
pressure
293 K, 6.3 Bars)
Quantity
Released
(pounds)
1,000
• 2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2^00
5,000
7,500
10,000
25,000
50,000
100,000
180,000
Quantity
In Pool
(pounds)
977
2,448
5,000
7,321
9,768
24,476
46,526
97,682
175,739
650
1,678
3,352
5,027
6,483
16,229
32,414
64,827
116,865
986
2,470
4,917
7,387
9,856
24,6%
49,172
98,564
177,282
858
2,227
5,005
6,659
8,577
21,433 '
42,777
85,775
154,350
Pool Fire
Radius
(meters)
11
15
19
23
25
36
46
60
74
9
13
17
20
21
30
39
51
64
10
15
19
22
25
35
45
58
73
10
14
18
21
23
32
42
54
67
Distance in Meters to the Following
Heat Radiation Levels t
1.6
kW/m2
112
151
190
217
238
321
403
506
614
100
136
170
194
211
284
356
447
542
114
153
192
219
241
324
407
511
620
110
150
188
214
232
313
393
493
598
4.0
KW/m1
72
%
120
137
150
203
255
320
388
63
86
108
123
. 133
180
225
283
343
72
97
121
139
152
205
257
323
392
70
95
119
135
147
198
249
312
378
12.5
KW/m2
40
54
68
77
85
115
144
181
220
36
49
61
69
75
102
128
160
194
41
55
69
78
86
116
146
183
222
39
54
67
77
83
112
141
176
214
Pool Fire
Duration
(seconds)
26
32
38
42
45
57
68
81
94
22
28
33
36
39
49
58
69
80
25
31
37
41
44
55
65
78
90
23
29
35
38
41
51
61
73
84
-------
-49-
Exhibit 16 (continued)
DISTANCES FOR POOL FIRES
FOR INSTANTANEOUS RELEASES OF 1,000 TO 180,000 POUNDS
DETERMINED USING WHAZAN
Distance in Meters to Three Heat Radiation Levels
Chemical
(Initial
Conditions)
Pentane
(293 K,
2.74 Bars)
Gasoline
(293 K,
1.013 Bars)
Hexane
(293 K,
1.013 Bars)
Heptane
(293 K,
1.013 Bars)
Quantity
Released
(pounds)
1,000
2,500
5,000
- 7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
Quantity
In Pool
(pounds)
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
Pool Fire
Radius
(meters)
10
14
19
22
24
34
45
58
72
10
14
18
22
24
34
44
57
71
11
15
19
22
25
35
46
59
74
11
15 •
20
23
26
36
47
' 61
76
Distance in Meters to the Following
Heat Radiation Levels
1.6
kW/m2
114
153
192
219
240
324
406
510
618
109
146
184
209
230
310
389
488
592
108
146
183
209
229
309
388
488
592
104
140
175
200
220
297
373
469
569
4.0
kW/m2
72
97
121
138
152
205
257
322
391
69
93
116
132
145
1%
246
309
375 -
69
92
116
132
145
196
246
308
374
66
88
11
127
139
188
236
297
360
12.S
kW/m2
41
55
69
78
86
116
145
182
221
39
52
66
75
82
111
139
175
212
39
52
65
75
82
111
139
174
212
37
50
63
72
79
106
134
168
204
Pool Fire
Duration
(seconds)
25
31
37
41
44
55
66
78
91
26
33
39
43
46
58
69
82
95
27
34
40
44
48
60
71
85
98
29
36
43
48
52
65
77
92
106
-------
-50-
Exhibit 16 (continued)
DISTANCES FOR POOL FIRES
FOR INSTANTANEOUS RELEASES OF 1,000 TO 180,000 POUNDS
DETERMINED USING WHAZAN
Distance in Meters to Three Heat Radiation Levels
Chemical
(Initial
Conditions)
Toluene
(293 K,
1.013 Bars)
p-Xylene
(293 K,
1.013 Bars)
Quantity
Released
(pounds)
1,000
2,500
5,000
• 7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
Quantity
In Pool
(pounds)
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
1,000
2,500
5,000
7,500
10,000
25,000
50,000
100,000
180,000
Pool Fire
Radius
(meters)
11
15
20
23
26
36
47
60
75
11
15
20
23
26
37
47
61
76
Distance in Meters to the Following
Heat Radiation Levels
1.6
kW/m2
95
128
160
183
201
272
338
429
520
94
127
159
181
199
269
338
425
517
4.0
kW/m2
60
81
101
116
127
172
214
271
329
59
80
101
115
126
170
214
269
327
I2g5
KW/m*
34
46
57
65
72
97
121
153
186
34
45
57
65
71
96
121
152
185
Pool Fire
Duration
(seconds)
32
40
48
53
57
71
87
100
117
33
41
49
54
58
73
87
103
119
-------
Exhibit 17
QUANTITY RELEASED VERSUS DISTANCE
FOR POOL FIRES, INSTANTANEOUS RELEASES
WHAZAN Modelling for 1,000-50,000 Pounds, 12.5 kW/m2
150
125
100
•
**
• '
u
n
75
50
25
-1-
f-
A
t *
• A
Etnyiene
**
Propylene
*
Propane
n-Butano
Pentana
• Gasoline
Hexane
10
20 30 40
Quantity Released (thousands of pounds)
50
-------
-52-
9.6 Jet Fire Results
Exhibit 18 presents WHAZAN results for jet fires, which are defined by the model as prolonged
releases. Results were obtained only for two gases (propylene and n-butane) assumed to be stored under
pressure at ambient temperature. Results for releases of 10,000 pounds and three heat radiation levels are
shown along with the duration of the jet fire. Distances for 10,000 pound releases of n-butane were less
than 25 meters for a all three heat radiation levels modeled. For propylene, the distances ranged from 35
meters at 37.5 kW/m2 to 51 meters at 4.0 kW/m2. The duration of the release (and presumably of the jet
fire) would be approximately one hour for the n-butane jet fire and about 15 minutes for the propylene jet
fire. For both of these gases, the other types of accidents modeled gave much greater distances than jet
fires. As detailed in Exhibit D-l, Appendix D, jet fires for 100,000 pound releases of n-Wane and
propylene produce distance results almost identical to the results for 10,000 pound releases. However, the
duration of the jet fires for the 100,000 pound releases was much greater. Note that for an instantaneous
release of these gases, the possible result would be a vapor cloud explosion or fire, not a jet fire.
10.0 Findings
WHAZAN model results seemed to closely correlate with overpressure data from actual vapor
cloud explosions. This helps to support the use of the WHAZAN model to predict consequences of vapor
cloud explosions. Based on the results of modeling flammable hazards using the WHAZAN model and
analysis of the literature, a clear distinction can be made between the results for flammable substances that
boil at higher temperatures than pentane and flammable substances that boil at lower temperatures than
pentane. This would suggest a distinction exists at roughly 310°K - 315°K (37°C - 42°C). Analysis of other
classification schemes and regulations (see Section 3) shows that flammable gases and volatile flammable
liquids appear to be the flammable materials considered of greatest concern by agencies and organizations
such as DOT, NFPA, and EEC. The temperature at which clear distinctions in consequences can be made
corresponds to the temperature at which DOT and NFPA have made distinctions in categories of
flammable materials (boiling point below 38°C and flash point below 23°C for NFPA flammability rating 4;
boiling point below 35°C and flash point below 23°C for DOT Packing Group I). This temperature also.
corresponds to very high ambient temperatures (38°C is roughly 100°F); it is possible for substances with
boiling points less than 37°C - 42°C to be gases under ambient conditions. Therefore, flammable gases and
very volatile flammable liquids (including substances that boil at temperatures less than about 42°C)
appear to be the most likely candidates for further consideration.
Exhibit 19 identifies 69 flammable gases (boiling point at or below 20°C). This list includes most
flammable gases currently in commerce. Eight of these substances are regulated as extremely hazardous
substances (EHSs) on the basis of toxicity and are indicated in the exhibit by an asterisk. The substances
in this exhibit are gases that have NFPA flammability ratings of 4 (NFPA 1984) or are listed by DOT as
flammable gases in current DOT regulations (U.S. DOT 1984) or in DOTs proposed rule (U.S. DOT
1987). (Note that there are discrepancies in the ratings of a few chemicals, e.g., vinyl bromide is listed as a
flammable gas by DOT but considered non-flammable by the NFPA.) Liquids that boil close to ambient
temperature may vaporize readily and also may form vapor clouds. If the flash points of such liquids are
at or below ambient temperature, the vapor may ignite under ambient conditions and vapor cloud fires or
explosions may occur. Exhibit 20 shows 28 liquids that boil below 38°C and have flash points lower than
23°C (i.e., their flash points are at or below normal ambient temperature). Three of them (indicated by an
asterisk in the exhibit) are listed as EHSs because of toxicity. The substances in this exhibit have NFPA
flammability ratings of 4 (NFPA 1984) or appear to meet the criteria for this rating on the basis of flash
point and boiling point.
-------
-53-
Exhibit 18
DISTANCES FOR JET FIRES
FOR PROLONGED RELEASES OF 10,000 POUNDS
DETERMINED USING WHAZAN
Chemical
(Initial
Conditions)
Propylene
(liquefied under
pressure
293 K, 13.3 Bars)
n-Butane
(liquefied under
pressure
293 K, 6.3 Bars)
Quantity
Released
(pounds)
10,000
10,000
Distance in Meters to the Following Heat
Radiation Levels
4.0 kW/m2
51
25
12.5 kW/m2
40
19
37.5 kW/m2
35
17
Jet Fire
Duration
^seconds)
15
60
-------
-54-
Exhibit 19
FLAMMABLE GASES
CAS #
74-86-2
7784-42-1
598-73-2
75-63-8
106-99-0
75-28-5
106-97-8
590-18-1
624-64-6
106-98-9
107-01-7
25167-67-3
630-08-0
463-58-1
7791-21-1
460-19-5
506-77-4
287-23-0
75-19-4
7782-39-0
19287-45-7
4109-96-0
75-68-3
75-37-6
124-40-3
463-82-1
74-84-0
107-00-6
75-04-7
75-00-3
74-85-1
75-21-8
353-36-6
540-67-0
109-95-5
50-00-0
7782-65-2
1333-74-0
7783-07-5
7783-06-4
Chemical
ACETYLENE
ARSINE*
BROMOTRI FLUORETHYLENE
BROMOTRI FLUOROMETHANE
1,3 -BUTADIENE
ISOBUTANE
BUTANE
2-BUTENE-CIS
2-BUTENE-TRANS
ALPHA -BUTYLENE
BETA-BUTYLENE
BUTYLENE
CARBON MONOXIDE
CARBON OXYSULFIDE
CHLORINE MONOXIDE
CYANOGEN
CYANOGEN CHLORIDE
CYCLOBUTANE
CYCLOPROPANE
DEUTERIUM
DIBORANE*
DICHLOROSILANE
D I FLUORO - 1 - CHLOROETHANE
DIFLUOROETHANE
DIMETHYLAMINE
2,2- DIMETHYLPROPANE
ETHANE
ETHYL ACETYLENE
ETHYLAMINE
ETHYL CHLORIDE
ETHYLENE
ETHYLENE OXIDE*
ETHYL FLUORIDE
ETHYL METHYL ETHER
ETHYL NITRITE
FORMALDEHYDE*
GERMANE •
HYDROGEN
HYDROGEN SELENIDE*
HYDROGEN SULFIDE*
Boiling Point (°C)
-83
-63
-59
-4.4 *
-12
-0.5
3.7
0.9
-6.5
1.0
-6.3
-192
-50
4
-21
13
13
-33
-250
-93
8.3
r.9
-25
7
9.5
-89
17
12
-104
11
-38
11
17
-20
-88
-253
-41
-60
EHS
-------
-55-
Exhibit 19
FLAMMABLE GASES (continued)
CAS #
64741-48-6
68476-85-7
74-82-8
74-89-5
563-45-1
74-87-3
115-10-6
593-53-3
74-93-1
115-11-7
8006-14-2
68476-26-6
504-60-9
7803-51-2
463-49-0
74-98-6
115-07-1
74-99-7
7803-62-5
116-14-3
79-38-9
420-46-2
75-50-3
689-97-4
593-60-2
75-01-4
75-02-5
75-38-7
107-25-5
Chemical Boiline Point (°C)
LIQUEFIED NATURAL GAS
LIQUEFIED PETROLEUM GAS
METHANE
METHYLAMINE
3- METHYL- 1-BUTENE
METHYL CHLORIDE
METHYL ETHER
METHYL FLUORIDE
METHYL MERCAPTAN*
2-METHYLPROPENE
NATURAL GAS
OIL GAS
1,3-PENTADIENE
PHOSPHINE*
PROPADIENE
PROPANE
PROPYLENE
PROPYNE
SILANE
TETRAFLUOROETHYLENE
TRI FLUOROCHLOROETHYLENE
1,1, 1 -TRIFLUOROETHANE
TRIMETHYLAMINE
VINYL ACETYLENE
VINYL BROMIDE
VINYL CHLORIDE
VINYL FLUORIDE
VINYLIDENE FLUORIDE
VINYL METHYL ETHER
-159
-40
-162
-6.3
20
-24
-24
6
-7
-43
-88
-35
-42
-48
-23
-111
-76
-28
-48 -
3
5
16
-13
-72
-83
12
EHS
Sources: NFPA 1984, U.S. DOT 1984, U.S. DOT 1987.
-------
-56-
Exhibit 20
FLAMMABLE, VOLATILE LIQUIDS
CAS #
75-07-0
627-20-3
646-04-8
75-91-2
503-17-3
557-98-2
590-21-6
75-18-3
60-29-7
75-08-1
110-00-9
74-90-8
78-79-5
563-46-2
563-45-1
107-31-3
78-78-4
109-66-0
109-67-1
8030-30-6
75-31-0
75-29-6
75-56-9
75-76-3
10025-78-2
109-92-2
75-35-4
Chemical Boiling Point
ACETALDEHYDE
AMYLENE, beta-.cis
AMYLENE, beta-, trans
TERT- BUTYL HYDROPEROXIDE
2-BUTYNE
2-CHLOROPROPYLENE
.1-CHLOROPROPYLENE
'DIMETHYL SULFIDE
ETHYL ETHER
ETHYL MERCAPTAN
FURAN*
HYDROGEN CYANIDE*
ISOPRENE
ISOPROPENYL ACETYLENE
2 -METHYL- 1-BUTENE
3 -METHYL- 1-BUTENE
METHYL FORMATE
ISOPENTANE
PENTANE
1-PENTENE
PETROLEUM NAPHTHA
ISOPROPYLAMINE
ISOPROPYL CHLORIDE
PROPYLENE OXIDE*
TETRAMETHYLSILANE
TRICHLOROSILANE
VINYL ETHYL ETHER
VINYLIDENE CHLORIDE
(°C)
20
37
36
35
27
23
33
37
35
34
32
26
34
33
31
37
32
28
36
30
35
32
36
34
26
32
36
32
Flash Point
ro
-36
<-20
<-fo
<27
<-20
<-20
<-6
-38
-45
-18
-18
-54
<-7
<-7
<-7
-32
-57
-49
-18
-18
-26
-32
-37
-14
-46
-28
*EHS
Sources: NFPA 1984, U.S. DOT 1984, U.S. DOT 1987.
-------
-57-
Chemicals with higher boiling points (38°C and above) with flash points at or below ambient
temperature will not vaporize readily under ambient conditions, but may ignite and burn. For these
chemicals, pool fires are a more likely event than vapor cloud fires and explosions. Modeling indicates
that pool fires are less hazardous than vapor cloud fires and explosions, based on the distance results.
Over 350 liquids with boiling points of 38°C or higher and flash points below 23°C have been identified,
including many common, high-volume substances such as gasoline, benzene, toluene, and the xylenes.
For vapor cloud explosions, BLEVEs, and pool fires, the distance for a given hazard criterion level
is proportional to the cube root of the quantity released (i.e, doubling the quantity of chemical used in the
analysis does not double the distance, but increases it by a factor of about 1.26, the cube root of 2). For
vapor cloud fires, the distance is not directly proportional to quantity, the exact relationship is not known,
but the quantity versus distance curve (see Exhibit 13) is similar in shape to the curves showing the cube
root relationship (see Exhibits 10, 15, and 17).
For a given quantity of a flammable chemical, distances for all types of accidents depend on the
hazard criterion level cHosen for use in the analysis; i.e, the overpressure level (for vapor cloud
explosions), concentration (for vapor cloud fires), or heat radiation level (for BLEVEs, pool fires, and jet
fires). The greatest distances calculated using the WHAZAN model resulted from vapor cloud explosions
and vapor cloud fires of gases and very volatile liquids. Vapor cloud explosions have been described as
unlikely for clouds containing less than 10,000 pounds of flammable gas (API 1990). As noted earlier, the
effects of vapor cloud fires are likely to be limited to a much smaller cross-sectional area than the effects
of vapor cloud explosions, since persons would have to be in the path of the engulfing fire for fatalities to
occur. The vapor cloud explosion results vary depending on the overpressure considered; at overpressures
of 3.0 psi and higher, modeling showed greater distances for BLEVEs than for vapor cloud explosions.
Pool fires appear to produce more localized consequences than vapor cloud explosions, vapor cloud fires,
and BLEVEs.
The consequence analysis results do not take into account the likelihood that a particular type of
consequence, such as a vapor cloud explosion, will result from a particular accident The specific
circumstances surrounding an accident (e.g., amount of material involved; release type, such as storage,
processing, transfer, transport) and the specific conditions (e.g., time of day, local meteorology and climate,
proximity and type of population) at the time of an accident may have significant effects on the severity
and range of consequences. Modeling cannot take all circumstances and conditions into account.
The modeling also indicates that meteorological conditions can have a major effect on the
consequences of vapor cloud fires, based on the dispersion of the flammable cloud, which impacts both
distance travelled and concentration. Storage temperatures of liquefied gases may have a significant effect
on consequence results for gas releases; at extremely cold storage temperatures, much less of the gas will
flash on release. Based on the results for prolonged releases under the conditions modeled, it appears that
instantaneous releases have consequences at greater distances than prolonged releases.
The greatest distance overall calculated using the WHAZAN model for an instantaneous release
of 10,000 pounds was for a vapor cloud explosion of propylene liquefied under pressure. The distance for
an overpressure of 0.3 psi was 800 meters; however, this overpressure is probably too low to cause serious
injury. For an overpressure of 1.0 psi, modeling produced a distance of 300 meters for a 10,000 pound
release of propylene. Note, however, that a 10,000 pound release under the conditions modeled would
produce a cloud containing approximately 3,300 pounds of propylene, a quantity that would have a low
probability of exploding (Prugh 1987). Vapor cloud fire results for a 10,000 pound release of ethylene,
based on dispersion to a concentration equal to 50 percent of the LFL, indicated a downwind distance of
550 meters, which is greater than the vapor cloud explosion distance of 300 meters for 1.0 psi; however,
the area within which the vapor cloud would be flammable (and hence fatal to anyone in the fire zone) is
-------
-58-
References
AlChE. 1987. American Institute of Chemical Engineers. Methods for Calculation of Fire and Explosions
Hazards. New York: AIChE
API. 1990. American Petroleum Institute. Management of Process Hazards. Production and Refining
Departments. API Recommended Practice 750. Washington, D.C: API
ASTM. 1969. American Society for Testing and Materials. Standard Method of Test for Flash Point of
Volatile Flammable Materials by Tag Open-Cup Apparatus. t
Brasie W., Simpson, D. 1968. "Guidelines For Estimating Damage Explosion," AICHE Symposium on
Loss Prevention in the Process Industries, February 18-23,1968.
CCPS. 1989. Center for Chemical Process Safety of the American Institute of Chemical Engineers. New
York: American Institute of Chemical Engineers.
EEC. 1982. European Economic Community. Official Journal of the European Communities, Volume
25, L230. Council Directive of 24 June 1982 on the major accident hazards of certain industrial activities.
Eichler T., Napadensky, H. 1977. "Accidental Vapor Phase Explosions on Transportation Routes Near
Nuclear Plants," IIT Research Institute, Prepared for Argonbe National Laboratory. April 1977.
Eisenberg NA et al. 1975. Vulnerability Model. A Simulation System for Assessing Damage Resulting
from Marine Spills. NTIS AD-A015-245. Springfield, VA.
FEMA, DOT, EPA. 1989. Handbook of Chemical Hazard Analysis Procedures. Federal Emergency
Management Agency, U.S. Department of Transportation, U.S. Environmental Protection Agency.
Grelecki C. "Fundamentals of Fire and Explosion Hazards Evaluation." Hazards Research Corp., AICHE
Today Series.
KJetz TA. 1977. Unconfined Vapor Cloud Explosions. Eleventh Loss Prevention Symposium, sponsored
by AIChE.
Lees FP. 1980. Loss Prevention in the Process Industries, Vol. 1. London: Butterworths.
Lenoir EM, Davenport JA. 1993. A Survey of Vapor Cloud Explosions: Second Update. Process Safety
Progress, Vol 12, no.l, January, 1993, p. 11
Lewis DJ. 1980. Unconfined Vapor-Cloud Explosions-Historical Perspective and Predictive Method
Based on Incident Records. Prog. Energy Comb. Sci., Vol. 6. Great Britain: Pergamon Press Ltd.
McRae T. 1984. The Effects of Large Scale LNG/Water RPT Explosions," McRae, T.G., et.al., Lawrence
Livermore National Laboratory. April 27, 1984.
M & M Protection Consultants Data base. 1990. "Large Property Damage Losses in the Hydrocarbon-
Chemical Industries: A Thirty-Year Review", 13th Edition, 1990.
Mudan KS. 1984. Thermal Radiation Hazards from Hydrocarbon Pool Fires. Prog. Energy. Combust.
Sci. 1984, Vol. 10. Great Britain: Pergamon Press Ltd.
-------
-59-
References (continued)
MHIDAS Database. 1993. Major Hazard Incidents Data Service. Developed by Safety and Reliability
Consultants of AEA Technology and UK Health and Safety Executive. Part of Silver Platter CD-ROM
database.
Nazario NN. 1988. Preventing or Surviving Explosions. Chemical Engineering, August 15, 1988.
NFPA. "Fire Protection Handbook," 16th edition, National Fire Protection Association (NFPA).
»
NFPA. 1984. National Fire Protection Association. Fire Protection Guide on Hazardous Materials, 8th
ed. Quincy, MA: NFPA.
NJEPDE. 1991. New Jersey Department of Environmental Protection and Energy. ACTOR Model
Thermal Energy Analysis Subroutine.
OSHA. 1990. Occupational Safety and Health Administration. Accident Inspection Report Database
Pineau JP, Chaineaux J, Lefin Y, Mavrothalassitis G. 1991. Learning from Critical Analysis of Hazard
Studies and from Accidents in France. International Conference and Workshop on Modeling and
Mitigating the Consequences of Accidental Releases of Hazardous Materials, New Orleans, LA, May 20-
24, 1991.
Prugh RW. 1987. Evaluation of Unconfined Vapor Cloud Explosion Hazards, International Conference
on Vapor Cloud Modeling, Cambridge, MA. November 2-4, 1987.
Resources for the Future. 1992. Fatal Hazardous Materials Database.
Richmond D. 1968. The Relationship Between Selected Blast-Wave Parameters and the Response of
Mammals Exposed to Air Blast" Richmond DR, Damon EG, Fletcher ER, Bowen IG, White CS., Ann.
N.Y. Acad. Sci., 1968.
Sadee, C. 1977. "The Characteristics of the Explosion of Cyclohexane at the Nypro (UK) Flixborough
Plant on 1st June 1974." Sadee C, Samuels DE, O'Brien TP, Journal of Occupational Accidents, 1977, 1:
p203-235.
Sax NI, Lewis RJ. 1987. Hawle/s Condensed Chemical Dictionary, llth ed. New York: Van Nostrand
Reinhold Co.
Skarka J. 1987. Considerations of Maximum Hazard Limits Originating from LPC Processing and
Handling, International Conference on Vapor Cloud Modeling, Cambridge, MA. November 2-4, 1987.
EPA, FEMA, DOT. 1987. Technical Guidance for Hazards Analysis, Emergency Planning for Extremely
Hazardous Substances. U.S. Environmental Protection Agency, Federal Emergency Management Agency,
U.S. Department of Transportation.
TNO. 1980. Bureau for Industrial Safety (TNO). Methods for the Calculation of the Physical Effects of
the Escape of Dangerous Material (the "Yellow Book"). Rijswijk, Netherlands: TNO (Commissioned by:
Directorate-General of Labour).
-------
-60-
References (continued)
U.S. Air Force 1983. "Explosives Safety Standards," AF Regulation 127-100, 20 May 1983.
U.S. DOT 1984. Hazardous Materials Table. 49 CFR 172,101.
U.S. DOT 1987. Proposed Rulemaking. 52 FR 42787-42931, November 6, 1987.
U.S. DOT 1988. "Hazard Analysis of Commercial Space Transportation," Office of Commercial Space
Transportation Licensing Programs Division, U.S. Department of Transportation. May 1988.
U.S. EPA. 1988. U.S. Environmental Protection Agency. Acute hazardous events data base (1988).
Draft final report. Prepared by Industrial Economics, Inc. EPA Contract No. 68-W8-0038; and disk
version of database. . .
U.S. EPA. 1992. Accidental Release Information Program (ARIP) Database.
van den Berg, AC. 1985. The multi-energy method, a framework for vapour cloud explosion blast
prediction. Journal of Hazardous Materials, 12 (1985) 1-10.
van den Berg, AC, van Wingerden, CJM, Zeeuwen, JP, Pasman, HJ. 1991. Current research at TNO on
vapor cloud explosion modeling. International Conference and Workshop on Modeling and Mitigating the
Consequences of Accidental Releases of Hazardous Materials, New Orleans, LA, May 20-24, 1991.
WHAZAN. 1988. WHAZAN User Guide. Technica International Ltd.
Wiekema BJ. 1984. Vapor Cloud Explosions - An Analysis Based on Accidents, Parts I and II, Journal of
Hazardous Materials, 8 (1984).
World Bank. 1985. Manual of Industrial Hazard Assessment Techniques. Office of Environmental and
Scientific Affairs, The World Bank.
-------
-A-l-
APPENDIX A
ADDITIONAL INFORMATION ON CLASSIFICATION SYSTEMS
FOR FLAMMABLE CHEMICALS
A.1 Definitions Related to Classification of Flammable Substances
A flammable material is "any solid, liquid, vapor or gas that will ignite easily and burn rapidly.
Flammable gases are ignited very easily; the flame and heat propagation rate is so great as to resemble an
explosion, especially if the gas is confined. Flammable gases are extremely dangerous fire hazards and
require precisely regulated storage conditions" (Sax and Lewis 1987).
The flash point of a chemical is "the temperature at which a liquid or volatile solid gives off a vapor
sufficient to form an ignitable mixture with the air near the surface of the liquid or within the test vessel"
(NFPA as cited by Sax and Lewis 1987). There are many different methods to measure a chemical's flash
point. Examples are the Cleveland Open Cup (COC) and Pensky-Martens methods. Most flash point
tests pass a flame just above the surface of the material being tested. The material is heated slowly and
the flash point is defined as the temperature at which the vapor generated above the material ignites.
Note that because ignition is the test end point, flash point tests can be used as a criterion for determining
whether a material is flammable or not. Flash point is also related to a chemical's volatility.
Unfortunately, most flash point tests are relatively inaccurate, particularly when applied to viscous liquids
(their precision is generally ±3°F (ASTM 1969)).
The boiling point of a chemical is "the temperature of a liquid at which its vapor pressure is equal
to or very slightly greater than the atmospheric pressure of the environment" (Sax and Lewis 1987).
Boiling point is an indicator of a chemical's volatility. Boiling point tests are relatively accurate and
simple to run.
A.2 DOT Classifications
Current DOT classifications for flammable materials are as follows (49 CFR 172):
Flammable Liquid: Any liquid having a flash point below 100°F (37.8°C).
Combustible Liquid: Liquid with flash point of 1008F (37.8°C) or greater, and less than 200°F
(93°C).
Flammable Solid: Any solid material, other than an explosive, which under normal transportation
conditions is liable to cause fires through friction or retained heat, or which
can be ignited readily and burns so vigorously and persistently as to create a
serious transportation hazard.
Flammable Gas: Compressed gas is defined as a material having pressure greater than 40 psia at
70°F, or pressure greater than 104 psia at 130°F, or a flammable liquid, vapor
pressure greater than 40 psia at 100°F; a compressed gas is classed as
flammable if mixture of 13% or less by volume with air forms flammable
mixture or flammable range is wider than 12%, or it is shown to be explosive
or flammable by one of several test procedures.
DOT published a proposed rule on November 6, 1987 to reclassify chemicals consistent with United
Nations (UN) International standards. The proposed flammable classifications are:
Hazard Class 2.1 Gases, ignitable when in a mixture of 13 percent or less by volume with air, or have a
flammable range with air of at least 12 percent regardless of the lower flammable limit. A gas is defined
-------
-A-2-
as a substance with a normal boiling point of 20°C (68°F) or less at one atmosphere pressure. Flammable
gases are not assigned to packing groups.
Hazard Class 3 Liquids, with flash points of 60.5°C (141°F) or lower. Packing groups are assigned
according to flash point and boiling point.
Hazard Glass 3 Packing Groups
Packing Group I - Boiling point less than 35°C (95°F).
Packing Group II -- Flash point less than 23°C (73°F) and boiling point greater than 35°C (95°F).
Packing Group III -- Flash point between 23°C (73°F) and 60.5°C (171°F) and boiling point
greater than 35°C (95°F).
A.3 NFPA Classifications
NFPA ratings for flammability are as follows (NFPA 1984):
NFPA 4 Assigned to materials that will burn readily and are readily dispersed in air or will
vaporize rapidly or completely at atmospheric pressure and normal ambient
temperature. Gases and cryogenic materials are included, as well as liquids or
liquefied gases with flash points below 22.8°C (73°F) and boiling points below 37.8°C
(100°F).
NFPA 3 Assigned to liquids and solids that can be ignited under almost all ambient
temperature conditions. Liquids with flash points below 22.8°C (73°F) and boiling
points at or above 37.8°C (100°F), or flash points at or above 22.8°C (73°F) and below
37.8°C (100°F).
NFPA 2 Assigned to materials that must be moderately heated or exposed to relatively high.
ambient temperatures before ignition can occur. Would not under normal conditions
form hazardous atmospheres with air. Liquids with flash points greater than 37.8°C
(100°F) and less than or equal to 93.4"C (200°F).
NFPA 1 Assigned to materials that must be preheated before ignition can occur. Considerable
preheating required under all ambient temperature conditions. Liquids, solids, and
semi-solids with flash point greater than 93.4°C (200°F).
NFPA 0 Assigned to materials that will not burn.
The classifications in NFPA 30 are:
Flammable Liquid Flash Point < 100°F (37.8°C)
Vapor Pressure < 40 psia @ 100°F (37.8°C)
Class IA Flash Point < 73°F (22.8°C)
Boiling Point < 100°F (37.8°C)
Class IB Flash Point < 73°F (22.8°C)
Boiling Point * 100°F (37.8°C)
Class 1C Flash Point > 73°F (22.8°C) and < 100°F (37.8°C)
-------
-A-3-
Combustible Liquid Hash Point * 100°F (37.8°C)
Class II 100°F (37.8°C) * Flash Point < 140°F (60°C)
Class IIIA 140°F (60°C) * Flash Point < 200°F (93°C)
Class IIIB Flash Point > 200°F (93°C)
NFPA 704 gives Class IA liquids a flammability rating of 4 and Class IB and 1C liquids a flammability
rating of 3.
A.4 EEC Indicative Criteria f
The Indicative Criteria related to flammability are as follows (EEC 1982):
Flammable Gases: Substances which in the gaseous state at normal pressure and mixed with air
become flammable and the boiling point of which at normal pressure is 20°C
or below.
Highly Flammable Substances which have a flash point lower than 21°C and the boiling point of
Liquids: which at normal pressure is above 20°C.
Flammable Liquids: Substances which have a flash point lower than 55°C and which remain liquid
under pressure, where particular processing conditions, such as high pressure
and high temperature, may create major-accident hazards.
A.5 OSHA Regulations for Service Stations
Some of the specific requirements for flammable liquids (e.g., gasoline) included in OSHA's
regulations for service stations are the following:
• Aboveground tanks in a bulk plant may be connected by piping to service station
underground tanks if a valve is installed within control of service station personnel;
• Flammable liquids can be dispersed from tank vehicles in non-public areas if the
vehicle, hose, and nozzle meet requirements;
• Class I (flammable) liquids cannot be stored in a basement or pit unless there is
ventilation;
• Electrical and heating equipment must meet certain standards when installed in areas
where flammable liquids are stored or handled;
• No smoking or open flames in areas used for fueling or servicing, and motors must be
shut off during fueling.
-------
-B-l-
APPENDIX B
METHODOLOGIES FOR MODELLING
B.I WHAZAN (World Bank Hazard Analysis)
B.I.I General Description
WHAZAN consists of a series of 13 consequence models (along with a database of chemical properties for
a selection of important hazardous chemicals) which can be run individually or linked together. The
consequence models consist of 4 release/evaporation models, 5 dispersion models, and 4 models dealing
with fires and explosions. For this analysis, the linked model for instantaneous and prolon|ed releases was
used. This model produces outputs which include pool fires, jet fires, vapor cloud explosions,
fireballs/BLEVEs, evaporation rates for pools, and flashed fractions for liquified gases. The evaporation
rate was input into the individual (stand-alone) vapor cloud dispersion model to determine the maximum
distance to 50 percent of the LFL. Similarly, flashed fraction was input into the vapor cloud explosion
stand-alone model to determine distances for vapor cloud explosions of liquified gases. The equations
used in this report are described below. The models can be run on an IBM PC and most likely on other
compatible systems.
B.1.2 WHAZAN Pressurized Release Outflow
For liquid stored in a refrigerated or pressurized container, WHAZAN calculates liquid outflow using the
Bernoulli equation. The equation assumes the driving force for outflow may be the liquid head or
difference between the storage pressure and the atmospheric pressure.
W = CDARDL[2(PrPA)/DL + 2gH]*
where:
W = release rate (kg/sec)
CD = coefficient of discharge
AR = area of release (m2)
DL = density of the liquid (kg/m3)
Pj = initial (storage) pressure (N/m2)
PA = atmospheric pressure (N/m2)
g = gravitational acceleration (m/s2)
H = liquid head (m)
The fraction of liquid that flashes once the fluid has reached atmospheric pressure is given by:
F— f"1 /T T ^ fU
v ~ M'tAM " 'B/'^vap
where:
Fv = fraction flashed to vapor
CPL = liquid specific heat (J/kg-K)
Tj = storage temperature (K)
TB = boiling point (K)
= heat of evaporation (J/kg)
-------
-B-2-
B.1.3 WHAZAN Two-Phase Outflow
For a two-phase outflow (liquid and gas) of a flashing liquid, the release rate of the liquid is given by:
W = ARCD (2Dm(Ps-Pc))*
where:
Dm = density of two-phase mixture (kg/m3)
Ps = saturated pressure (N/m2)
Pc = choked pressure (N/m2)
The density of the 2-phase mixture at the exit plane is given by:
Dm = 1/(F^DV) + (1-FV)(DL)
where:
Dv = density of the vapor (kg/m3)
The fraction of the release that has flashed to vapor at the exit plane, F^, is given by:
where:
CPL = Liquid specific heat (J/kg-K)
Ts = Saturation Temperature (K)
B.1.4 WHAZAN Gas Outflow
These equations may be applied to the discharge of toxic and flammable gases from large vessels or pipes.
This model assumes reversible adiabatic expansion and ideal gas behavior. To calculate gas outflow rates,
WHAZAN uses the following:
W = YCDARP1[MG(2/(G+l))(G+1)/(°-1VRT]1'2
where
W = gas outflow (kg/s)
Y = coefficient in gas outflow model
CD = coefficient of discharge
AR = area of release (m2)
P! = initial (storage) pressure (N/m2)
M = molecular weight
G = ratio of vapor specific heats at constant volume and at constant pressure
R = universal gas constant (N-m/K-mole)
T = temperature (K)
In most cases of interest, flow will be critical because upstream pressure exceeds:
In these cases Y = 1.0. In the event that the upstream pressure is less than that given by the above
expression, then:
Y = (PA/P1)1/G[l-(PA/Pi)(G-1)/G]1/2[2((G+l)/2)(G+1)/(°-1V(G-l)]1/2
-------
-B-3-
B.1.5 WHAZAN Adiabatic Expansion Outflow
These equations describe the initial behavior of an instantaneous pressurized release. The outflow may be
used for subsequent dispersion calculations. The model assumes a core of uniform concentration
containing 50% of the released mass surrounded by a peripheral zone characterized by a Gaussian
distribution of concentration. For the estimation of the rapid adiabatic expansion experienced during the
instantaneous release of a flashing liquid or pressurized vapor, WHAZAN uses a simple two zone
hemispherical model. The expansion process is considered to have two stages:
(1) In the first stage, gas expands down to atmospheric pressure or liquid flashes (as
appropriate).
The change in internal energy is:
»
UrU2 = C/IVTz)
The energy of expansion is:
E =
where
Uj = initial internal energy (J/kg)
U2 = final internal energy (J/kg)
Cy = vapor specific heat at constant volume (J/kg-K)
Tj = storage temperature (K)
T2 = temperature after initial expansion (K)
E = energy of expansion (J)
PA = atmospheric pressure (N/m2)
Vl = initial volume (m3)
V2 = final volume (m3)
(2) In the second stage, the kinetic energy developed by the initial expansion drives the turbulent
mixing of air into the cloud as the cloud spreads outward.
Once the expansion energy, E, has been determined, the expression for the turbulent coefficient is:
KD = 0.0137E1/2VG01/3[tE*/VG01/3]~1/4
and the expression for the core radius as a function of time is:
rc = 1.36(4KDt)1/2
where
KD = turbulent diffusion coefficient (m2/sec)
E = expansion energy (J)
VGO - volume of a unit gas at standard conditions (kg/m3)
t = time (sec)
rc = radius of cloud core (m)
The equation for the concentration in the core is:
jc = 0.0478VGO/((4KDt)3/2)
-------
-B-4-
The following are expressions for the core concentration and radius at the end of the mixing:
jc = . 172.95E-0'9
rc = 0.08837Ea3VG01/3
It was found (Ulden 1974) that the end of the initial expansion occurred at a cloud radius, r, such that r/rc
= 1.456. At this value of r/rc, 91% of the released material is still within the cloud.
B.I,6 WHAZAN Evaporation from Liquid Pool
For an instantaneous spill, if the atmospheric temperature is less than the normal boiling point of the
liquid, evaporation is calculated by:
—
dt
where
dm
dt = change of mass with time (kg/sec)
Ps = saturated pressure (N/m2)
M = molecular weight
M = molecular weight
R = universal gas constant (N-m/K-mole)
TA = atmospheric temperature (K)
U = wind speed at 10 m height (m/s)
r = radius of pool (m)
4.785x10 (neutral atmospheric stability)
0.25 (the midpoint of the observed range)
a
n
If the atmospheric temperature > normal boiling point of the chemical, two alternative models
apply. If the dominant means of transfer of heat to the pool is wind, the previous equation is appropriate.
If the dominant heat transfer mechanism is conduction through the ground, the appropriate model is as
follows:
where
k = 6.68xl05 to represent an average soil
B.1.7 WHAZAN Pool Fires
To describe a pool fire phenomenon, WHAZAN uses equations governing pool spread and burning rate.
In pool spread, the liquid is assumed to form a circular pool of uniform height. For an instantaneous
spill, the pool radius is given by:
r = (t/S)* (m)
-------
-B-5-
where:
E = (7rDL/8 gm)y'
where:
DL = density of liquid (kg/m3)
g = gravitational acceleration = 9.8 m/s2
m = released mass in cloud/pool (kg)
For continuous spills:
r = (t/g)*
where:
fi = (97rDL/32gW)1/3
To calculate burning rate for liquids having boiling points above ambient temperature:
dm 0.001 Hc
where:
Hc = heat of combustion (J/kg)
CPL = liquid specific heat (J/kg-K)
H^p = heat of evaporation (J/kg)
TB = boiling point (K)
TA = atmospheric temperature (K)
dm/dt = pool spread (kg/sec)
For liquids having boiling points below ambient temperature:
0.001Jfc
The heat released is calculated as:
(72
at
-------
-B-6-
where:
r = radius of pool (m)
The efficiency factor tj has been shown by tests to have a value up to around 0.35.
The flame height is calculated as:
H = 84 r
10.61
1/2
where:
g = gravitational acceleration (m/s2)
Numeric input data requirements for the "instantaneous" case are as follows:
Parameter Valid Range Default
Spill size 10 - 107kg
Bund area 0 - 106m^ 0
Ambient temperature 200 - 400°K 68°F
Efficiency factor 0.01 - 1 0.35
For the continuous case, the spill size is replaced by the spill rate and duration.
B.1.8 WHAZAN Jet Fires
The flame length is given by:
L= 18.5 W°-41
where:
W = release rate (kg/s)
L = length (m)
This method should be applied when the release is of a flashing liquid. When the release is essentially
gaseous, an alternative method is recommended. In this case, the flame length is based on the distance to
the lower flammability limit (LFL) as given by the jet dispersion model. To calculate jet dispersion, the
envelope within which the concentration of a turbulent jet is above a specified level, jc, is represented by a
major axis of length:
-------
-B-7-
A =
and a minor axis of length:
B = A
T1W
The shape parameters are given by
bj = 50.5 + 48.2 Dva - 9.95Dva2
b2 = 23.0 + 41.0 Dva
where:
DJ
= gas density at ambient conditions, relative to density of air (kg/m3)
= gas density at stated conditions, relative to density of air (kg/m3)
= specified level of a turbulent jet - the envelope of concentrations of a turbulent jet is above
this level (kg/m3)
= diameter of jet once expanded to atmospheric pressure (m), prior to entrainment of air
The jet shape is symmetrical about the major axis. D. is calculated assuming adiabatic expansion.
B.1.9 WHAZAN Vapor Cloud Explosions (based on Buoyant Plume Dispersion)
WHAZAN model expresses vapor cloud explosions in terms of overpressure versus distance.
Overpressurization is related to the dispersion concentration of the gas and assumptions about its
explosive yield. WHAZAN uses the standard Gaussian model of plume dispersion to calculate the
ground-level concentration of a buoyant release (where the release point can be at ground-level or at
height, h):
where:
W = release rate (kg/sec)
y = crosswind distance (m)
-------
-B-8-
h = release height (m)
U = wind speed at 10 m height (m/sec)
cr
-------
-B-9-
B.2 ARCHIE (Automated Resource for Chemical Hazard Incident Evaluation)
B.2.1 General Description
In a combined effort, FEMA, U.S. DOT, and U.S. EPA sponsored the development of a handbook and
computer program entitled ARCHIE to provide emergency planning personnel with the tools necessary to
evaluate the nature and magnitude of chemical release threats from potentially hazardous facilities.
A core part of the program estimates the downwind dispersion of a chemical release given a diversity of
release scenarios. The program operates on an IBM personal computer or other compatible system.
ARCHIE is capable of addressing a wide variety of common accident scenarios. The scenarios include
nine methods to estimate discharge rate and duration of a gas or liquid release from a tank or pipeline,
seven methods to estimate size of the liquid pool, two methods to estimate the rate at whi^h a liquid pool
will evaporate or boil, seven methods dealing with explosions and fires, and one method to estimate
downwind chemical concentrations and hazard zones for the dispersion of vapor clouds.
B.2.2 ARCHIE Pressurized Liquid Release
For a given liquid height and vapor space pressure, the instantaneous liquid release rate from a tank is
given by the equation:
m = AhCd Jp,[2gplHL - fg + 2(P0 - />„)]
where:
m = Discharge rate (kg/s)
g = Gravitational constant = 9.8 m/s2
p, = Liquid density (kg/m3)
P0 = Storage pressure (N/m2)
Pa = Ambient pressure (N/m )
HL = Liquid height above bottom of container (m)
Hh = Height of discharge opening (m)
Aj, = Area of discharge opening (m2)
Cd = Discharge coefficient
This assumes the tank is full. An average release rate for the tank at atmospheric pressure may be
obtained by computing the time (Te) required to empty a tank. For a spherical tank:
T.-
15 D02 fa Cd
where:
DT = Tank diameter (m)
Do = Opening diameter (m)
Te = Time to empty (sec)
The above equation assumes liquid is released through a circular opening at the bottom of the tank.
-------
-B-10-
B.2.3 ARCHIE Estimating Pool Size
Estimating pool areas that result from discharges of liquids is difficult and error-prone. Therefore, this
model is composed of several methods that provide results that are approximate, but reasonable.
For non-boiling liquids, a pool area may be calculated based on user-supplied data from observations made
at the potential accident site. The user may select use of the maximum credible pool area, or may choose
a simple and very crude correlations based on experimental data, this being:
log(A) = 0.492 log(m) + 1.617
where:
m = total liquid mass spilled (Ibs)
A = Pool area (ft2)
To compute pool areas for boiling liquids, one first needs to calculate the vaporization flux, E^ which is
then used in the pool spreading model. The vaporization flux is given as:
Ev = Fp
92.6 EXP(-O.OQ43T^
x 10
'7
where:
Ey = vaporization flux (kg/m2/sec)
F = 0.5322-0.001035Tb
Tb = boiling point (°F)
p = liquid density (kg/m3)
= molecular weight (kg/kmol)
£ = liquid specific gravity
This factor is used to determine the diameter of the pool fire.
Pool fire diameters can be calculated for continuous spills, where the spill continues at a specified finite
rate for a long duration; instantaneous spills, which occur in a very short time; and finite duration spills,
where a given volume of liquid is spilled over a given time interval. Pool fire diameters can also be
calculated for surfaces with friction, and with no friction.
The maximum diameter of a pool is calculated for a sample scenario-an instantaneous release of a boiling
liquid in the absence of friction:
3 I"8
A... « 1.7766 K A
where:
Dmax = maximum diameter of pool (m)
V = total liquid volume
A = effective gravity (same as gravitational constant, g, for spills) (m/sec2)
Ey = vaporization flux (kg/m2/sec)
-------
-B-ll-
The time needed to reach the maximum diameter, tmax, is given as:
11/4
= 0.6966
AE.
B.2.4 ARCHIE Evaporation of Liquid Pool
Assuming the discharged liquid is near ambient temperature, a simplified model developed by the U.S. Air
Force Engineering and Services Laboratory is used to predict the evaporation rate. The evaporative flux is
given as:
E = 4.66xlO-6C/075 "
where:
= Evaporation flux (lbs/min/ft2)
Uw = Wind speed (miles/hr)
Ps = Vapor pressure of chemical (mm Hg)
Psh = Vapor pressure of hydrazine (mm Hg)
MV, = Molecular weight of chemical
TF = Spill temperature correction factor
The spill temperature correction factor is defined as follows:
TF = 1 T < 0°C 2
TF = 1 + 4.3 x lO'^p Tp > 0°C
where Tp is the pool temperature in degrees C. The vapor pressure of hydrazine is given by the following
eauatinn:
equation:
ln(F) = 65.3319 - . . g.221n(7) + 6.1557 JcKT3 T
where T is in kelvins and P is in atmospheres.
Overall evaporation rate:
where
Vw = evaporation rate (kg/s)
Ey = vaporization flux (kg/nr/s)
A = pool area (m2)
-------
-B-12-
B.2.5 ARCHIE Gas Discharge from Pressurized Vessel
For the gas discharge model, the initial rate from a pressurized vessel is also calculated. The model
assumes that the process is adiabatic, and that wall friction is negligible. An expression for an
instantaneous discharge rate under non-choked flow conditions is given as:
m =Ah
Y-l
Under choked flow conditions, the mass flow rate is calculated from:
m
2 \^
2 Y-,
l»/2
f—1
U+iJ
where:
m = Discharge rate (kg/s)
Ah = Opening area (nr)
y = Ratio of specific heats
p0 = Tank pressure (Pascals)
Pj = Ambient pressure (Pascals)
P0 = Density (kg/m3)
This equation is based on ideal gas behavior.
B.2.6 ARCHIE Vapor Cloud Fire Model
The purpose of this model is to estimate the dimensions of the downwind area that may be subjected to
flammable and potentially explosive vapors and gases in the event of an accidental discharge. It applies
both to toxic gases and vapor cloud fires. The model also estimates the maximum weight of flammable gas
that may be airborne at any time. The size of the dispersion zone depends on the quantity of material
released, its effective density, volatilization, prevailing atmospheric conditions, source elevation, and user-
specified toxicity limit.
Either a neutrally buoyant or heavy gas model could be used to predict cloud dimensions. For neutrally
buoyant gases, ARCHIE considers releases as point sources at ground level. The emission rate is constant.
Ambient temperature was assumed to be 20°C. The model assumes steady state unless the release
duration is significantly smaller than the characteristic downwind travel time. Relatively short-duration
releases were assumed to be instantaneous. ARCHIE uses the finite duration model validated by Palazzi
to predict gas dispersion. According to the model, the finite concentration (Cf) at any location is given by:
C -
c' T
when t s.
-------
-B-13-
erf
-erf
x-Ut
(2)^0,
where:
C = centerline concentration given as:
C =
C 2*0,0,17,.
exp -
- — + exp -
2o.2 2o2
Here, tR is the duration of release in seconds. The maximum concentration is given by the following
equation:
Cc
ma. ~ "T~
C =
mu A
The parameters used in the above equations are defined as follows:
Cmax = Maximum centerline concentration (kg/m3)
Q = Continuous source release rate (kg/sec)
Uw = Wind speed (m/sec)
z = Vertical distance (m)
y = Crosswind distance (m)
x = Downwind distance (m)
H = Source height (m)
c;x = Longitudinal standard deviation (m)
o = Lateral standard deviation (m)
oz = Vertical standard deviation (m)
The dispersion distances for neutrally buoyant gases are generally 3 to 5 times larger than those for heavy
gases.
-------
-B-14-
B.2.7 ARCHIE Unconlined Vapor Cloud Explosion
ARCHIE uses the TNT-equivalent model for use in evaluating the vapor cloud explosion scenario. The
amount of combustion energy in the cloud is compared with the equivalent amount of TNT. Considering
that only a fraction of the energy in the cloud will contribute to the explosion (this is known as the yield
factor), and referring to the TNT explosion/overpressurization data, one can use the model to calculate the
pressurization at distances from the explosion. The fraction of energy in the cloud assumed to contribute
to the explosion ranges from 2 to 20 percent. Other simplifying assumptions include:
• Ambient temperature is 20°C;
• Effects of terrain, buildings, obstacles have not been considered.
The equation is stated as: .
m
TJfT
Atf
m... x —— x Y.
"cloud
1155
where:
mTNT = TNT equivalent mass (Ibs)
AHC = Lower heat of combustion (kcal/kg)
mcloud = Mass m cloud (lbs)
Yf = Yield factor
Distance to a given overpressurization is then calculated from the equation:
X = m^ exp (3.5031 - 0.7241 In (Op + 0.0398
where:
X = distance to given overpressure (ft)
Op = Peak overpressure (psi)
B.2.8 ARCHIE Tank Overpressurization Explosion Model
This model assumes that the pressurization waves created by an exploding tank will propagate omni-
directionally in a hemispherical field at ground level. The computational algorithm proceeds as follows:
1. Calculate the ratio Pj/Pa where Pa is the absolute ambient pressure and Pj is the absolute
internal gas pressure at which the tank is expected to rupture.
2. Compute the ratio T/Ta where Ta is the absolute ambient air temperature and Tj is the
absolute temperature of the gas in the tank.
3. Determine the initial overpressure ratio, P^, by solving the following equation by trial and
error:
-------
-B-15-
/ = 0 = In A - ln(l «• PJ - -i- In
" 1
2|F
where 7 is the ratio of specific heat at constant pressure to that at constant volume.
4. Compute the nondimensional starting distance R0 from:
1
4*
Y, - 1
1/3
5. Compute the value of R from:
R =
1/3
where:
V = Volume of the gas in the tank (ft3)
r = Distance from the center of the tank at which the side-on overpressure is desired (ft)
6. Locate the point associated with P^ and R0 on an overpressurization graph.
7. Follow the nearest curve for Ps vs. R to the R value computed in step 5. Read the Ps value
associated with this R value. If the gas vessel is on the ground and/or close to a reflecting
surface, increase Ps by 100% for R less than 1 and by 10% for R greater than 1.
8. The side-on overpressure is determined by multiplying the above resultant value of Ps by the
absolute value ambient atmospheric pressure Pa.
B.2.9 ARCHIE Fireball Model
In calculating the maximum diameter and height the fireball attains, as well as the safe separation
distances for fatality and injury, ARCHIE makes the following assumptions:
-------
-B-16-
Fuel is propane or has similar characteristics
Ambient temperature is 20°C
Atmospheric absorption of thermal radiation is negligible
Fraction of combustion energy radiated = 0.2
Observer is at ground level
Minimum fatality zone is equal to half the maximum diameter calculated
Based on a series of experiments:
Z = 26.3W1/3
T = 2.23W1/6
where:
W = Mass in vessel (Ibs)
Dmax = Maximum diameter of fireball (ft)
Z = Maximum height of fireball (ft)
T = Duration of fireball (s)
The safe separation distance for fatality, XF, in feet, was found to be:
XF = 1.48W0-56 W a: 2000 Ibs
XF = 8.0W0-33 W < 2000 Ibs
For injury, the safe separation distance, XI, in feet, is:
XI = 4.53W0-52
B.2.10 ARCHIE Liquid Pool Fire Model
ARCHIE calculates various pool fire factors to determine the radius in which injuries and fatalities are
expected from the fire. These factors include burning velocity, maximum pool diameter, flame height,
effective emissive power, incident flux, and view factor (fraction of flame seen by a given observer). In
calculating these factors, there are several simplifying assumptions:
Pool area is circular
Observer is at ground level
Ambient temperature is 20°C
Atmospheric absorption of thermal radiation is negligible
Negligible wind in the vicinity of the flame; thus, uniform thermal radiation field radially and
no flame tilt
Pool ignites shortly after release
Burning rate equals spill rate
The scenario chosen to model is an instantaneous liquid hydrocarbon release in the absence of frictional
resistance during spreading. The equation to estimate the burning velocity is:
MW
P6
-------
-B-17-
where:
y = Burning velocity (m/s)
MW = Molecular weight (kg/kmol)
p = Liquid specific gravity
TB = Normal boiling point (°F)
In the pool fire scenario, the radius of the pool increases until all the material is consumed by the fire.
The maximum diameter and time to reach maximum diameter are given by:
= 1-7766
K3A
1/8
= 0.6966
Ay2
where:
V = Total liquid volume (m3)
y = Burning velocity (m/s)
A = Effective gravity (same as gravitational constant (g) for spills on land) (m/s2)
The time averaged pool diameter is obtained by dividing the maximum diameter by the square root of two.
The mean visible flame height is based on equations correlated with data from laboratory fires. Visible
flame height, Hflaine, is expressed as:
BV p
10.61
where:
Hflame = F13"16 height
p = Liquid density (kg/m )
pa = Air density at ambient temperature (kg/m3)
Dp = Pool diameter (m)
g = Gravitational acceleration = 9.8 m/s2
Effective emissive power of the flame accounts for the incident flux shielding by surrounding layers of
smoke for liquid hydrocarbon fires. Based on literature data and correlated to the normal boiling point,
effective emissive power is defined as:
Ep = -0.313 TB + 117
-------
-B-18-
where:
Ep = Effective emissive power (kW/m2)
TB = Normal boiling point (°F)
The incident flux at any given location is given by the equation:
Qincident = Ep X r X VF
where:
Qincident = Incident flux (kW/m2)
T = Transmissivity t
VF — Geometric view factor
T, the transmissivity coefficient, is mainly a function of the path-length (distance from observer to flame
surface), relative humidity, and the flame temperature. For the calculation scheme in ARCHIE, r has
been set to 1, and the attenuation of thermal flux due to atmospheric absorption is not taken into account.
This assumption provides a conservative hazard estimate, since the presence of water and carbon dioxide
tends to reduce the incident flux at any given location.
The view factor defines the fraction of flame that is seen by a given observer. This geometric term has
been calculated as a function of distance from the flame center for an upright flame approximated by a
cylinder. It has also been assumed that the optimum orientation between observer and flame that yields a
maximum view factor prevails. The resulting equation is as follows:
where:
X = Distance from flame center (m)
Rp = Pool radius (m)
For fatality, the incident flux level is set to 10 kW/m2. For injury, the corresponding level is 5 kW/m2.
These levels are based on analysis of numerous sources of experimental burn data (Mudan, 1984).
Applying these two damage criteria, the above equations were rearranged to solve for hazard distances X10
and X^ for fatality and injury, respectively:
v =0.30— -
10 0.3048
0.43 r— E,OSJ
0.3048 '
-------
-B-19-
where:
X10 = Radius for expected fatalities (ft)
Xo5 = Radius for expected injuries (ft)
-------
-B-20-
B.3 Yellow Book
B.3.1 General Description
The Yellow Book includes calculations for pool fires, vapor cloud fires, and vapor cloud explosions for
gases; and pool fires and vapor cloud fires for liquids. For gases, two different methodologies are used for
vapor cloud fires: the flash fraction of each gas is modeled as an instantaneous release and the non-
flashed fraction is modeled as a continuous release from an evaporating pool. In both cases, dispersion as
a neutrally buoyant cloud is assumed. The flash fraction of each gas is used to calculate the vapor cloud
explosion results. Since there is no calculated flash fraction for liquids, no vapor cloud explosion results
are calculated for liquids.
B.3.2 Vapor Cloud Explosion Calculations for Gases
Reference: Yellow Book, Chapter 4, "Spray Release," p. 44; Chapter 7, "Dispersion," pp. 25-27, and
Appendix 2, p. 52; Chapter 8, "Vapor Cloud Explosion," pp. 24-25.
For gases, the flash fraction for a spray release is calculated. This fraction is then multiplied by the total
quantity to obtain the quantity instantaneously in the air. The maximum fraction of this quantity that is
between the upper and lower flammabilily limits is calculated. Damage circles for the resulting quantity
can then be determined.
Equations:
Mass Flashed (Chapter 4, p. 44)
W = X • W
vap ^ap,a s
where
Wvap = mass flashed (kg)
pa = weight fraction of vapor after expansion (flash fraction)
Ws = mass stored (kg)
The flash fraction is calculated from the following equation (Chapter 4, p. 16, equation (3)):
Tb TbC. T.
_ V ™ PI 1_ 1
where:
= weight fraction of vapor before expansion (assumed to be 0 for calculation of the flash
fraction)
Tb = boiling temperature of gas compressed to liquid (K)
Tj = temperature of stored gas compressed to liquid (K)
Cj = specific heat of gas compressed to liquid (J/kg-K)
hy = heat of evaporation of gas compressed to liquid (J/kg)
Flash fractions for selected gases are listed in Chapter 4, p.44, Table 1.
-------
-B-21-
B.3.2.1 Quantity of Vapor Contributing to Explosion (Chapter 7, pp. 25-26).
The equation in this section estimates the amount of gas in the explosive range. This amount develops as
the dispersion progresses. Initially, and close to the source, the concentration in a large part of the cloud
will be greater than the upper explosive limit; M^ (the amount of gas in the explosive range) will then be
small. As the dispersion progresses, M^ increases to a maximum concentration that is reached when the
maximum concentration is a little above the upper explosive limit. After this M^ will gradually drop
again.
= ERF M - ERF (fi) - EXP (-v2) + EXP (-
V* V"
where
M^ = amount of gas in explosive range (air not counted) (kg)
M = total amount of gas which has escaped (kg)
ERF(x) = error function =
X
— f EXP (-t*)dt
7
(These values are listed in Appendix 2 on p 52.)
where
v = Pj/Pj, v, = [Ln (vfl/fv2 - 1)
_ v2ln(v)
"
and
Pj = upper explosive limit (UEL)
P2 = lower explosive limit (LEL)
B 3.2.2 Determination of Damage Circles (Chapter 8, p. 25)
For the limit value of certain types of damage(s) the relation is expressed below:
-------
-B-22-
R(S) = the radius of a damage circle (m)
C(S) = constant for characteristic type of damage (mJ"1/3)
A constant C(S) goes with each characteristic damage (S), listed on p.25. For glass damage causing injury,
C(S) = .15 (a corresponding overpressure is not given).
Only part of the total combustion energy in the explosive part of the cloud is available for shock wave
propagation. This fraction is symbolized by the yield factor, n. This yield factor can be divided into two
independent factors ric and 7/m, in which 77 = t]c x r/m. i\c indicates the yield loss as a result of the non-
stoichiometry of a cloud with a continuous development of fuel concentration in the explosive part of the
cloud; it is put at 30%. ijm gives the mechanical yield of the combustion. Depending slightly on the type
of gas, i)m is calculated as follows: t
a. Isochoric combustion: rjm = approx. 33%
b. Isobaric combustion: rim = approx. 18%
Where a gas cloud explosion occurs, there is probably some form of confinement which would result
isochoric combustion and a probable value for tjm of 33%. 17 would then equal approximately 10%.
To determine E, the energy content of the explosive part of the vapor cloud, an equation is given
example on p. 26:
in
as an
E = mass x hc
where
E = energy content (J)
hc = heat of combustion (J/kg)
mass = explosive part of vapor cloud (kg)
B.3.3 Yellow Book Vapor Cloud Fire Calculations for Gases
Reference: Yellow Book, Chap. 5, "Evaporation," and Chap. 7, "Dispersion."
The evaporation rate for the pool left following the flash of an instantaneous release of a chemical is
calculated using the following:
W^p = mass flashed
The quantity remaining is:
Wpoo. = "'
where
Wpool = mass in P001 (k8)
= mass stored (kg)
The total volume of the chemical in the pool is calculated using the equation on page 42 of Chapter 5 in
the Yellow Book:
-------
-B-23-
where
Vl 0 = initial volume (m3)
WJ = mass in pool (kg)
P! = density (kg/m3)
The pool of liquid will spread out until the minimum layer thickness is reached. The time, te, at which
this takes place is best calculated using iterations of the equation on page 23:
8
3
(TE x e2 x C" x g x Vlfl x $
Pi2 J
16
"'-
-2
C" x g x KI
IJD
C"
0
where
Ts =
Tb =
as =
C" =
g =
a min
volume of the spreading pool (m3)
X, x (T, - Tb) / 1^ x (,r x as)1/2 [p. 16] (kg/in2-**)
temperature of substrate
boiling point of the liquid (K)
thermal diffusivity of substrate = 1.0 x 10"6 m2/s for concrete
constant for spreading liquids = 2
acceleration due to gravity = 9.8 m/s2
time for minimum pool thickness to be attained (to be iterated) (s)
density of the liquid (kg/m3)
= minimum thickness = 0.01 m
coefficient of heat conduction of substrate = 1.1 W/m-K
heat of evaporation of the liquid (J/kg)
This equation is used reiteritively to compute values of te until a minimum positive value is found.
The evaporation rate at the minimum pool thickness is calculated as follows:
Mp - 4 x (n x e1 x C" x g x Vlo x t^ - (8/3) x TC x C" x g x e x 8, ^ x *;
3/2
where
Mp = the instantaneous evaporation rate at the minimum pool thickness (kg/m2s)
and all other variables are as defined above.
A dispersion distance to 50% of the lower flammability limit is calculated using the equation on page 15
of the Dispersion section of the Yellow Book (where y, z, and h all equal 0):
-------
-B-24-
M.
[2 x n x Uw x oy(x) x at(x)]
where:
C(x, y, z) =
M
ff(X)
a
concentration at coordinate x, y, z = 50% of LFL (kg/m3)
evaporation rate at minimum pool thickness (as described above) (kg/sec)
axb = 0.128xa905 (m)
ex41 = 0.2x°-76 (m)
(using values in the table on page 10 for neutral conditions)
wind speed at 10 m height (m/sec)
B.3.4 Vapor Cloud Fire Calculations for Gases (Based on Fraction Flashed, 50% LFL)
Reference: Yellow Book, Chapter 4, "Spray Release," pp. 36, 44; Chapter 7, "Dispersion," p. 23.
The following equations estimate the consequence distance for a vapor cloud fire for gases, assuming that
a percentage of the total amount released is immediately vaporized; this vaporized quantity is assumed to
be available for the vapor cloud fire. The quantity evaporating from the pool that remains after flashing is
not considered in this calculation.
The flashed mass is calculated as follows (Chapter 4, p 16):
Wf = mass flashed (kg)
a = flash fraction (see Section B.3.2 for calculation)
p a
Ws = mass stored (kg)
The quantity Wf is then used as an instantaneous source using equation 23 (p. 24 in "Dispersion" section),
where the concentration at a point (x, y, z) at time t is given as:
C(x,y,z,t) =
m
(2
EXP
EXP
2o2
17
EXP
(z-K?
2'z,2
EXP
-------
-B-25-
where:
m = source strength for an instantaneous release (kg)
x = coordinate in wind direction (m)
y = coordinate at right angles to wind direction (m)
z = coordinate in vertical direction (m)
Uw = wind speed at 10 m height (m/s)
h = source height (m)
°Xl °\\ azi = standard deviations in x, y, and z directions for an instantaneous source
We assume that y, z, and h = 0, and that x = Uwt, where Uw is wind speed, to determine the distance the
cloud has traveled in the x direction only:
C(x,0,Q,t) = W
From p. 23, axi, aYi< and az[ are described by these equations:
-------
-B-26-
Solve for x:
y 11/2.665
,3/2
B.3.5 Pool Fire Calculations for Gases and Liquids
t
Reference: Chapter 4, "Spray Release," p. 44; Chapter 6, "Heat Radiation," pp. 8-14, 19-35.
To determine the heat radiation into the environment, the radiating surface (the flame from the burning
pool) is seen as an upright cylinder with diameter D and length L. The radiation load q per unit area
exposed which is experienced at a distance r from the center of the fire is (Ch. 6, p. 8):
qr = TJ x F x E
where
qr = radiation load per unit area at a distance r from the center of the fire (kW/m2)
TJ = atmospheric coefficient of transmission
F = geometric view factor
E = average intensity of radiation (kW/m2)
BJ.5.1 Atmospheric Coefficient of Transmission
Part of the radiated heat is absorbed by the air between the object exposed to radiation and the fire. This
reduction is taken into account by r}, the coefficient of transmission. The size of this coefficient is
dependent upon the amount of water vapor between the flame and the object, the air temperature, and the
relevant distance (r) between the object and the center of the fire.
The relevant distance (r) is the unknown variable for the purposes of this study. qr the radiation load, is
assumed to be 12.5 kw/m2, the intensity that corresponds to one percent probability of fatality for a 30
second exposure. Figure 2 on p. 35 gives T! as a function of the product of vapor pressure and relevant
distance. The vapor pressure is determined by multiplying the relative humidity by the saturated vapor
pressure (given in Table 1, p.19) at a prevailing temperature. To be conservative, a value of 10 percent
was used for relative humidity.
To estimate the relevant distance, an iterative process, including calculations and estimates of rl
(atmospheric coefficient of transmission) and F (geometric view factor) from B.3.5.1, B.3.5.2, and B.3.5.3,
was used. Assuming a relevant distance and multiplying that estimate by the vapor pressure yields an
estimate for t, using the table on page 35. This value for t can then be used in calculating the view factor
(F), as described in B.3.5.2 and B3.53. The estimate for F is used to determine another estimate for the
relevant distance. This new assumed distance is used to estimate another value for t, and the steps above
are reiterated. This reiterative process is used until the relevant distance approaches a single value.
, B.3.5.2 The Geometric View Factor F
The effect of the geometrical shape of the flame and the place and orientation of the exposed object are
incorporated in the view factor F. The size of the view factor is determined by the length to diameter
ratio of the flame (Chapter 6, p. 9).
-------
•B-27-
i-42
d
m
where
Lf = length of the flame (m)
d = diameter of the flame (m)
m" = rate of evaporation (kg/m2s)
g = acceleration due to gravity = 9.8 m/s2
pa = density of air (at 18°C = 1.213 kg/m3) t
The diameter of the flame is assumed equal to the diameter of the pool. The diameter of the pool is
calculated from the volume of propane in the pool, assuming the pool spreads out to a thickness of .01 m.
The calculation for the rate of evaporation (m") is dependent upon whether the stored material is a liquid
with a boiling point below the ambient temperature or above the ambient temperature.
Liquids with a boiling point below ambient temperature are gases stored under high pressure conditions.
The equation is given in Chapter 6, p. 10:
m" = — . 1(T3 *g/m2sec
where
hc = heat of combustion (J/kg)
hy = heat of vaporization (J/kg)
For liquids with a boiling point above ambient temperature (Ch. 6, p.ll):
m
. 1C
'3
where
C = specific heat or heat capacity at constant pressure (J/kg.K)
AT = boiling temperature - ambient temperature (K)
The value of the view factor (not the size) can be determined by solving the initial equation for radiation
load for F:
qr
F
x F x E
-------
-B-28-
B.3.5.3 Intensity of Radiation of the Fire
The value for E, the average intensity of radiation (W/m2), is dependent on both the type of fuel and on
the diameter of the pool. The magnitude of the average intensity of radiation is given on Table 3, p. 21
for four gases: butane, propane, ethylene, and propylene. The value for the other substances is calculated
using the equation in Chapter 6, Appendix 2, p. 32 for liquids with a boiling point above ambient
temperature:
£ =
0.35
C,
*«
AT +
103
*.
72
.61
where
E = radiation intensity (W/m2)
hc = heat of combustion (J/kg)
Cp = heat capacity (J/kg-K)
hy = heat of vaporization (J/kg)
AT = boiling point - ambient temperature (°K)
The length to diameter ratio can then be determined using the equation in Chapter 6, p. 9:
Table 1 on p. 24 gives view factors between a vertical upright cylinder and a surface on the ground. The
table lists the view factors as coordinates between the ratio of flame length to radius and the ratio of the
distance of the object from the fire to radius. These ratios are diagrammed in Appendix I. Using the
section of the table that gives maximum view factors, the ratio a/b is assumed to be the length to radius
ratio and is taken to be twice the flame length to diameter ratio. Using the known ratio a/b and the view
factor F, the ratio c/b can be estimated and the relevant distance (c) to the object receiving radiation can
be determined. Using this new assumed distance, a new value for rl can be estimated in a reiteration of
the process, until the distance approaches a single value.
B3.6 Vapor Cloud Fire for Liquids (Dispersion Distance to 50% LFL)
Reference: Yellow Book, Chapter 5, "Evaporation," Pp. 29-38; Chapter 7, "Dispersion," Pp. 15-16.
The calculations in this section model the consequence distance in a vapor cloud fire when the material
evaporates from a pool on the ground rather than flashing into a vapor state after an instantaneous
release. The consequence distance to a concentration that is 50% of the lower flammability limit (LFL) is
modeled. The 50% LFL figure is a conservative level that accounts for uncertainties like the uneven
dispersion within the cloud and the estimated value of the lower flammability limit To determine the
mass flux of evaporating non-boiling liquids, the following equation is used (Chapter 5, P. 33):
-------
-B-29-
2.1(T3 U°-(l r^/1 -^- (Pw - ?») for. Pw < 2x10* Pa
wt 111 poOifC n*T. " "
*v«_i.
where .
ft
mnb = evaporation rate of non-boiling liquids (kg/m2s)
Uw 10 = wind speed at 10 meters = 3 m/s
r | c = radius of confined liquid pool = 46.8 in
M = molecular weight (kg/kmol)
R = molar gas constant = 8.3x103 (J/kmol-K)
Tnb = temperature of non-boiling liquid = ambient temperature = 293 K
Pw = partial vapor pressure on liquid surface (Pa)
Poo = partial vapor pressure in surroundings = 0 for liquid (Pa)
In order to calculate the dispersion to 50% LFL, the equation for the concentration for a continuous
source on p. 15 of Chapter 7 is used:
C(x,y,z) =
with y, z, and h = 0, this equation reduces to:
m
where
C(x) = concentration as a function of x only (kg.m"3)
m = source strength (kg-s"1)
Uw = wind speed at 10 m height (m/s)
h = source height (m)
-------
-B-30-
B.4 AIChE-Sponsored Course Materials
B.4.1 General Description
The American Institute of Chemical Engineers (AIChE) course material includes calculations for pool
fires and vapor cloud fires for gases and liquids. For vapor cloud fires for gases, the flash fraction of each
gas is modeled as an instantaneous release; the AICHE material explains that the consequence distance
produced by the initial flash fraction cloud is so much larger than the consequence distance for the
remaining non-flashed part that it dominates in assessing the hazard area. The calculations for vapor
cloud explosions and BLEVES appear incomplete and cannot be used as methodologies.
B.4.2 Vapor Cloud Fire Calculation for Gases t
Reference: AIChE Pp. B-65 and B-85; D-9 - D-10.
The fraction of liquified gas vaporized (v) is given on p. D-9:
v = fraction of liquid vaporized
v = 1 - expiCp/h/T,, - T,)}
where
Cp = specific heat (Btu/lb°F)
hy = latent heat of vaporization (Btu/lb)
Tb — normal boiling point (°F)
Tj = temperature of liquid stream (°F)
We then multiply the fraction of the liquid vaporized by the quantity released to determine the amount of
vaporized material (Q).
To determine the distance to 50% LFL, the equation on p. B-84 is used:
x(0, o, o, t) = 131Q/(Ut)Z62 for neutral conditions
where
, o, o, t) = concentration at the center of the cloud (assume 50% LFL) at time t (kg/m3)
U ' ' = wind speed (m/s)
t = time (s)
Ut = distance (m)
The equation is solved for Ut, the distance, by including the 50% LFL concentration as X and the quantity
of vaporized material, Q.
-------
-B-31-
B.4.3 Vapor Cloud Fire Calculations for Liquids
Reference: AlChE Pp. D-2 - D-5, D-9 - D-1U.
To determine the rate of evaporation from a circular pool, the area of the pool must first be determined.
This is calculated by dividing the total volume of liquid by the thickness of the pool, which is estimated for
different surface types in Table 1 on p. D-4. The radius of the pool is then determined and used in the
following equation for evaporation (p. D-2):
E = K'U^-n)/^+n¥4+nw/+n) (kg/sec)
where
K> — a'., r»2n/(2+n) 7 -n/(2+n)
— a Mo u L\
Mo = psM _ vapor density (kg/m3)
RT
Ps = saturated vapor pressure (N/m2)
M = molecular weight (kg/kmol)
R = universal gas constant (8314.3 J/mol K)
T = ambient temperature, K
D = vapor diffusivity value
Values of n, a', KYU0 are given in Table II (using neutral stability).
The distance that a vapor cloud disperses from a liquid pool is given by the equation on p. D-9:
d! = {(36.8)(Q)/(U)(CL)}°-552 (m)
where
dj = distance at which cloud is diluted to LFL (m)
Q = source strength or volumetric vapor escape rate (m3/sec)
U = wind velocity (m/sec)
CL = lower explosion limit (voL^p^/volgj,.)
(used 50% LEL versus LEL)
The volumetric vapor escape rate (Q) is calculated from the Evaporation rate using the Ideal Gas Law.
The evaporation rate (kg/s) is converted to moles using the molecular weight, moles are converted to liters
at the prevailing temperature, and liters may then be converted to m3.
B.4.4 Pool Fire Calculations for Liquid and Gases
Reference: AIChE Pp. D-16 - D-29.
The AIChE methodology does not distinguish between liquids and gases in the calculation of the thermal
flux at a specified distance from a burning pool. To determine the liquid burning rate (v), the following
equation is used (P. D-17):
v = v, {l-exp(-kjd)} (for a pool of diameter d)
where
-kj = attenuation coefficient
v = liquid burning rate (Ibs/s)
-------
d
hc
h'
-B-32-
liquid burning rale of infinite pool = .0030 hp/h^, (in/min)
pool diameter (ft)
heat of combustion (Btu/lb)
sensible heat of vaporization (Btu/lb) = sensible heat + latent heat
The mass burning rate can then be calculated (p. D-17):
M - vp
where
M = mass burning rate (lb/ft2 sec)
v = liquid burning rate (in/min) t
p = density of the liquid (lb/ft3)
The mass burning rate is then used to estimate the ratio of flame height (h) to pool diameter (d) (p. D-
18):
m
1.61
where
m
pa
g
d
= mass burning rate (kg/m2-sec)
= density of ambient air (1.206 kg/m3)
= acceleration of gravity = 9.78 m/sec2
= pool diameter (m)
Thermal flux is calculated by dividing the total heat (Q) by the surface area of the flame envelope (S).
Total heat available during the pool fire is:
Q =
(Btu/sec)
where
hc = heat of combustion (Btu/lb)
m = mass burning rate (Ib/ft2-sec)
A = pool area (ft2)
Pool area is determined by dividing the total volume of the liquid by the thickness of the pool.
Thicknesses for spreading pools over different surfaces are listed in Table I, page D-4.
Only a fraction of the total heat is radiated from the surface of the flame envelope. These fractions may
be read from Table II, page D-5. The surface heat flux (q0) can then be determined (p. D-19).
q0 = Q/S
(Btu/ft2hr)
where
S = surface area of cylindrical flame envelope = 2irrh + irr2 (ft2)
Q = total heat (Btu/hr)
-------
19):
-B-33-
The heat absorbed (qr) by a receiver at various distances (x) from the pool center is given by (p. D-
Tx Fx
where
q0 = heat flux at flame surface (Btu/ft2 hr)
er = receiver surface absorptivity (from Table III, D-25 - assumed to oe 1, all heat absorbed)
T
X
atmospheric transmissivity at distance X = l-ew
Fx = view factor from receiver to flame
ew = emissivity of water in air
The atmospheric transmissivity is estimated as a function of distance (x) from the center of the flame to
the receiver times the partial pressure of water. ew is read from a graph of emissivity versus the partial
pressure of water (aim) times the distance (ft) found on p. D-26, assuming the partial pressure is .0024 for
worst case winter conditions.
The view factor (F) is read from tables on pp. D-27 - D-29. It is plotted as a function of distance (D)
from the flame centerline to the receiver flame radius and the plots are arranged by the flame height/pool
diameter ratio.
The distance is initially assumed to find a value for E^ and q^ is equal to 12.5 kw/m2. The equation is
then solved for Fx, and the distance "D" can be determined. Successive approximations of the calculations
for the value of Fx are then calculated using the new values for the distance until one value is approached.
-------
-B-34-
B.5 Green Book
The Technical Guidance for Hazards Analysis ("Green Book") contains calculations for the consequence
distances for Vapor Cloud Fires, where 50% of the lower flammability limit (LFL) or another fraction of
the LFL may be used as concentration level.
Reference: Green Book, Appendix G, "Equations Used for Estimation of Vulnerable Zones."
The rate of release of a chemical is needed for calculation of the distance for a specific concentration. It
is dependent on the quantity of chemical released, the nature of the release scenario (i.e., pool of liquid),
and the properties of the chemical released. For spilled pools of chemicals, the rate of release is usually
taken to be the evaporation rate (rate of volatilization). Using the assumptions presented above; the
following equation is used to calculate the rate of release to air for liquids (in Ibs/min):
R _ (60 sec/min xMWxKxAxVPx (929cm2/fr2)
R x (Tl +273) x (IQQmmHgltam) x 454gllb
where
QR = rate of release to air (Ibs/min)
MW = molecular weight (g/g mole)
K = gas phase mass transfer coefficient (cm/sec)
A = surface area of spilled material (ft2)
VP = vapor pressure of material at temperature Tl (mm Hg)
R = 82.05 aim cm3/g mole K
Tl = temperature at which the chemical is stored (°C)
The equation for the evaporation rate (rate of volatilization) can be rewritten as follows:
OR = °162 xMWx
K(.T1 +273)
K can be estimated based on a known value for a reference compound as follows:
K = K^ x (MWref/MW)1/3
Using water as the reference compound:
Knrf - Kwaier = 0-25 X (U)0'78
where: U = windspeed (m/sec)
Combining the two equations above:
K =0.25(U)°-78x(18/MW)w
Combining equations for QR and K yields the following equations (Green Book p. G-2):
-------
-B-35-
0.162 x 0.25 x (IT)0-78 x (18)l/3 x MW213 x A x VP
Rx (17+273)
_ 0.106 * (tfl0-78 x MW213 xAxVP
Rx(Tl +273)
Calculation of the surface area (A) of the spilled material is carried out as described below.
The following assumptions are used to calculate the surface area of the spill:
Density = 62.4 lb/ft3 (i.e., all liquids are assumed to have the same density as water)
Depth of pool is 0.033 ft (1 cm)
The surface area of the spilled liquid (ft2) is (Green Book p. G-3):
A- =0.49xQS
62AU>lfi3 x Q.033ft
where
QS = quantity spilled (Ibs)
A = surface area (ft2)
Substituting for A in the above equation for release rate, the quantity released to air per minute (QR) can
be estimated as follows:
QR = 0-106 x (IQ07» x MW3 x 0.49 x QS x VP
82.05 x (77 +273)
For gases, it is assumed (Green Book p. 3-3) that the entire mass is released in the ten minutes:
QS (Ibs)
QR =
10 minutes
where
QS = quantity spilled (Ibs)
QR = rate of release to air (Ibs/min)
-------
-B-36-
The following equation was used to derive the distance to a concentration (C). The concentration
downwind of a release is given by (Green Book p. G-4):
QR
for a ground level release with no effective plume risk where:
C ss airborne concentration (g/m3) (assumed 50% LFL)
QR = rate of release to air (g/sec)
TT = 3.141
ffy x
-------
•C-l-
APPENDIX C
HAZARD CRITERIA
C.I Explosion Overpressure
C.I.I Damage and Injury from Blast Waves
Blast waves from accidental explosions can cause damage to people and property by subjecting
them to transient crushing pressures and winds. Relatively simple concepts have been used to quite
effectively correlate blast wave properties with damage. The concept states that damage is primarily a
function of either the peak overpressure, the impulse or some combination of these factors. Peak
overpressure is most commonly used. *
Criteria for peak overpressures causing personal injury are given in Exhibit C-l, including injury
from direct blast effects and from flying glass. Guidelines for peak overpressures required to produce
property damage are presented in Exhibit C-2. These data are largely based on empirical observations.
There seems to be geneYal agreement between sources on the data presented in Exhibit C-2; however, the
same is not true for the data presented in Exhibit C-l. According to Exhibit C-l, lung damage may occur
at overpressures of 10 psi. According to other sources, threshold lung damage may not occur until peak
overpressures reach 30 to 40 psi (Grelecki, Richmond 1968). The values for corresponding fatalities also
vary. According to the Exhibit, the lowest overpressure cited for fatality from direct blast effects is 14.5
psi. According to another source, a 1% mortality rate is probable at 27.0 psi. Others cite values in the
range of 35 to 120 psi as the threshold overpressure value for fatalities, 20.5 to 180 psi for a 50% fatality
rate, and 29 to 250 psi for a 99% fatality rate (Grelecki, Richmond 1968, Lees 1980).
Exhibits C-3 and C-4 show how lethality and injury vary with both overpressure and the duration
of the blast wave. Exhibit C-5 shows similar data as a function of both pounds of TNT and range in feet.
Exhibits C-3 through C-5 present potential injury levels resulting from direct exposure to blast
overpressures in an open area. They do not take into account injuries resulting from property damage
presented in Exhibit C-2. For example, substantial injury and possible deaths may result from the
shattering of concrete walls at 2.0 to 3.0 psi. In planning with regard to injury to personnel, the following
guidelines have been recommended (Brasie and Simpson 1968):
• Personnel in areas subject to overpressures greater than about 2 psi are
likely casualties from fragmentation or self-impact against objects.
• Personnel beyond the under 1 psi overpressure range should be
reasonably safe inside a reinforced structure away from windows or, if
outdoors, lying on the ground.
Nazario (1988) recommends that all personnel in open terrain be evacuated if the potential of 0.3
psi overpressure exists and recommends evacuating personnel in buildings that might experience an
overpressure exceeding 1.0 psi (0.25 psi overpressure if the building has windows).
According to the AIChE (1987), the usual procedure for control room design for plants where
there may be the potential for explosions is to design the control room for a peak overpressure on the
order of 1 to 3 psi.
-------
-C-2-
EXHIBITC-l. HUMAN INJURY CRITERIA
(Includes Injury from Flying Glass and Direct Overpressure Effects)
Overpressure
(psi)
Injury
Comments
Source
0.6 Threshold for injury
from flying glass*
1.0 - 2.0 Threshold for skin
laceration from flying glass
1.5 Threshold for multiple
skin penetrations from
flying glass (bare skin)*
2.0 - 3.0 Threshold for serious wounds
from flying glass
2.4 Threshold for eardrum rupture
2.8 .1.0% probability of eardrum
rupture
3.0 Overpressure will hurl
a person to the ground
3.4 1% eardrum rupture
4.0 - 5.0 Serious wounds from flying
glass near 50% probability
5.8 Threshold for body-wall
penetration from
flying glass (bare skin)*
6.3 50% probability of eardrum
rupture
Based on studies
using sheep and dogs
Based on Army data
Based on studies
using sheep and dogs
Based on Army data
Conflicting data on
eardrum rupture
Conflicting data on
eardrum rupture
One source suggested an
overpressure of 1.0 psi
for this effect
Not a serious lesion
Based on Army data
Based on studies
using sheep and dogs
Conflicting data on
eardrum rupture
(Fletcher, Richmond,
and^Yelverton 1980)
(Lees 1980)
(Fletcher, Richmond,
and Yelverton 1980)
(Lees 1980)
(Lees 1980)
(Lees 1980)
(Brasie and Simpson
1968)
(U.S. DOT 1988)
(Lees 1980)
(Fletcher, Richmond,
and Yelverton 1980)
(Lees 1980)
Interpretation of tables of data presented in reference.
-------
-C-3-
EXHIBITC-1. HUMAN INJURY CRITERIA
(Includes Injury from Flying Glass and Direct Overpressure Effects)
(continued)
Overpressure
(psi)
Injury
Comments
Source
7.0 - 8.0
10.0
14.5
16.0
17.5
20.5
25.5
27.0
Serious wounds from flying
glass near 100% probability
Threshold lung
hemorrhage
Fatality threshold for
direct blast effects
50% eardrum
rupture
10% probability of fatality
from direct blast effects
50% probability of fatality
from direct blast effects
90% probability of fatality
from direct blast effects
1% Mortality
.29.0
99% probability of fatality
from direct blast effects
Based on Army data
Not a serious lesion
(applies to a blast of
long duration (over 50
msec); 20-30 psi required
for 3 msec duration waves)
Fatality primarily from
lung hemorrhage
Some of the ear injuries
would be severe
Conflicting data on
mortality
Conflicting data on
mortality
Conflicting data on
mortality
A high incidence of severe
lung injuries (applies to a
blast of long duration (over
50 msec); 60-70 psi required
for 3 msec duration waves)
Conflicting data on
mortality
(Lees 1980)
(U.S. DOT 1988)
(Lees 1980)
(U.S'. DOT 1988)
(Lees 1980)
(Lees 1980)
(Lees 1980)
(U.S. DOT 1988)
(Lees 1980)
-------
-C-4-
EXHIBIT C-2. PROPERTY DAMAGE CRITERIA
Overpressure (psi)
Damage
Source(s)
0.03
0.04
0.10
0.15
0.30
0.4
0.5 - 1.0
0.7
1.0
1.0 - 2.0
1.3
Occasional breaking of large glass
windows already under strain
Loud noise (143dB). Sonic boom
glass failure
Breakage of windows, small, under strain
Typical pressure for glass failure
"Safe distance" (probability 0.95 no
serious damage beyond this value). Missile
limit. Some damage to house ceilings.
10% window glass broken.
Minor structural damage
Shattering of glass windows, occasional
damage to window frames. One source
reported glass failure at 0.147 psi
Minor damage to house structures
Partial demolition of houses, made
uninhabitable
Shattering of corrugated asbestos siding
Failure of corrugated aluminum/steel paneling
Failure of wood siding panels (standard
housing construction)
Steel frame of clad building slightly
distorted
(Lees 1980)
(Leej 1980)
(Lees 1980)
(Lees 1980)
(Lees 1980)
(McRae 1984, Lees
1980)
(Brasie and Simpson
1968, Air Force 1983,
U.S. DOT 1988,
Lees 1980)
(Lees 1980)
(Lees 1980)
(Brasie and Simpson
1968, Air Force 1983,
U.S. DOT 1988,
Lees 1980)
(Lees 1980)
2.0
Partial collapse of walls and roofs of houses (Lees 1980)
-------
-C-5-
EXHIBIT C-2. PROPERTY DAMAGE CRITERIA
(continued)
Overpressure (psi)
Damage
Source(s)
2.0 - 3.0
2.3
2.5
3.0
3.0 - 4.0
4.0
4.8
5.0
5.0 - 7.0
7.0
7.0 - 8.0
Shattering of non-reinforced concrete or cinder
block wall panels (1.5 psi according to
another source)
Lower limit of serious structural damage
50% destruction of brickwork of house
Steel frame building distorted and
pulled away from foundations
Collapse of self-framing steel panel buildings
Rupture of oil storage tanks
Snapping failure - wooden utility tanks
Cladding of light industrial buildings
ruptured
Failure of reinforced concrete structures
Snapping failure - wooden utility poles
Nearly complete destruction of houses
Loaded train wagons overturned
Shearing/flexure failure of brick wall panels
(8-12 inches thick, not reinforced)
Sides blown in of steel frame buildings
Overturning of loaded rail cars
(Brasie and Simpson
1968, Air Force 1983,
U.S. DOT 1988,
Lees 1980)
(Lees 1980)
(Lees 1980)
(Lees 1980)
(Brasie and Simpson
U.S. DOT 1988,
Lees 1980)
(Lees 1980)
(McRae 1984)
(Brasie and Simpson
1968, Lees 1980)
(Lees 1980)
(Lees 1980)
(Brasie and Simpson
1968, Air Force 1983,
U.S. DOT 1988,
Lees 1980)
(Air Force 1983)
(Brasie and Simpson
1968, U.S. DOT 1988)
-------
-C-6-
EXHIBIT C-2. PROPERTY DAMAGE CRITERIA
(continued)
Overpressure (psi)
Damage
Source(s)
9.0
10.0
30.0
Loaded train box-cars completely demolished
Probable total destruction of buildings
Steel towers blown down
(Lees 1980)
1980)
(Brasie and Simpson
1968, Air Force 1983)
88.0
Crater damage
(McRae 1984)
-------
-C-7-
EXHlBrrC-3. LETHALITY CURVES
(For a 154 Ib Penon in Free Stream Situations)
Mttbmuffl
tun
10.000
5.000
2.000
1.000
500
200
MJ
t.SOO
1.000
700
400
200
100
70
40
20
10
50%L0th«lity —-- —
IHUthUity -
•H-
OJ 0.4 1.0 2 * 710 20 *0 100 200 1000 5000
10*
101
tun
10
0.1
EXHIBIT C4. LETHALTTY AND IN JURY CURVES
(For • 154 Ib Perm In Fm Strt«« Stautkws)
1.000
100
10
0.1
0.01
lartfrum rupturt
tS% lartfrum pretKtion with muffl
95%
•*-
10
OJ 1.0
OurtMn flf poiittvt in
Notes: msec « mfllisecoods
kPt m kfloptscab « 6J895 pd
psi » pounds per square inch » 0.1450 kPa
Source: US. DOT 1988, KokmaHs and Rudolph 1981
100
1000
•4-
10.000
-------
-C-8-
EXfflBIT 05. AIR.BLAST CRITERIA FOR PERSONNEL STANDING IN THE OPEN
u.
o
en
o
1000
1000
INJURIES FROM =
OISPLACCMCNT
% or KMSONNCL
•LOWNOOWN
KMSONNCL
ftLOWNOOWN
1% MOKTAUTT MOM
OmtCT
TMKCSHOLO UM«
OAUAOI
100
RANGE IN FEET
1000
-; 0.001
iS
10,000
Sottree US. DOT 1988, Richmond and Fletcher 197L
-------
-C-9-
Note that although glass is reported to shatter at an overpressure level of 0.5 psi (see Exhibit C-
2), there is evidence that human injury from flying glass is unlikely at this level. Lees (1980) cites a report
stating that risk of injury from flying window glass is negligible for an explosion that gives a peak
overpressure of 0.6 psi or less. Lees also cites a report giving the skin laceration threshold for flying glass
as 1-2 psi and the serious wound threshold as 2-3 psi (see Exhibit C-l). Fletcher, Richmond, and
Yelverton (1980) carried out experiments on the effects on sheep and dogs of glass fragments from
windows shattered by various blast overpressures. Their results (briefly cited in Exhibit C-l) appear to
agree reasonably well with the data reported by Lees. Their data indicate that the threshold for injury
from flying glass is about 0.6 psi. Multiple injuries from skin penetration (i.e., ten or more) may occur
for bare skin at overpressures of 1.5 psi or higher. Hying glass may penetrate the body-wall at 5.8 psi or
higher if the skin is bare. The authors point out that the skin and body-wall thickness of men and sheep
are approximately the same (Fletcher, Richmond, and Yelverton 1980); therefore, the data reported should
be roughly applicable to humans.
C.I.2 Distance Determination for Different Overpressures
To determine the effect on distance of modeling to different overpressures for vapor cloud
explosion, the WHAZAN model was run for several different overpressures for vapor cloud explosions of
ethylene and propylene. Exhibit C-6 shows explosion overpressure versus consequence distance for
100,000 pound releases of ethylene and propylene. The curves are not linear; a relatively small change in
overpressure at the lower end of the scale results in a relatively large change in the consequence distance,
while a relatively small change in overpressure at the upper end of the scale results in relatively small
change in consequence distance. However, increasing the overpressure at the upper end of the
overpressure scale appears to have a much smaller effect on the consequence distance.
C.2 Flammability Limits
As noted in the text (Section 8.2), the lower flammable limit (LFL) of a vapor or gas is the
minimum concentration in air that will ignite and propagate flame. Using 50 percent of the LFL to
estimate consequence distances for vapor cloud fires accounts for variability within the cloud from its edge
to its center and inaccuracies in dispersion modeling.
To determine the effect of using different concentrations to estimate the consequence distance for
vapor cloud fires, two representative chemicals, ethylene and propylene, were modeled using WHAZAN at
10,000 pounds and 100,000 pounds. Moderate meteorology (wind speed 3.0 meters per second and
atmospheric stability D) was assumed. The model was run for three concentrations, at 50% of the lower
flammable limit, at the lower flammable limit, and at 200% of the lower flammable limit. The results of
modeling are shown graphically in Exhibit C-l. The results curves for both quantities and both chemicals
modeled are basically similar in shape but of differing magnitudes. The results curves clearly are not
linear: the change in magnitude between the consequence distance obtained at 50% of the lower
flammabile limit and that at the lower flammable limit is clearly greater than the change between the
lower flammable limit and 200% of the lower flammable limit for all four curves.
-------
Exhibit C6
HAZARD CRITERIA COMPARISON FOR VAPOR CLOUD EXPLOSIONS
FOR INSTANTANEOUS RELEASES OF ItMM POUNDS
1750
1600 —
1260 —
g
£
E
g 1000 —
600
260
o
9
760 —
4 56
Ov*rpr««sur« (p«l)
-------
Exhibit C-7
HAZARD CRITERIA COMPARISON FOR VAPOR CLOUD FIRES
FOR EVAPORATING POOLS FROM INSTANTANEOUS RELEASES, MODERATE METEOROLOGY
(Wind speed 3.0 meten per second, atmospheric stability class D)
1500
1250-
1000—
0) 750-
O
c
D
"w 500-
Q
250-
A
*
O
Elhylene (100.000 Iba.)
Propylene (100.000 Iba.)
Ethylene (10.000 Iba.)
Propylene (10.000 Iba.)
n
1
50% LFL
LFL
2LFL
Concentration
-------
-C-12-
C.3 Heat Radiation
C.3.1 Heat Radiation Levels Causing Death and Injury
Human injury or fatality from heat radiation from fires is likely to occur as a result of direct
exposure to a fire rather than as a result of property damage caused by a fire. The extent of the injury
caused by heat radiation depends both on the heat level and the time of exposure. Exhibit C-8 presents
heat radiation levels and exposure times required to produce various human effects or injuries. There are
few data on fatalities from heat radiation; the fatality data shown in Exhibit C-8 have been calculated or
estimated. Both the WHAZAN (1988) and Mudan (1984) fatality data appear to be based on data on the
relation between thermal radiation intensity and burn injury for nuclear explosions at different yields
(Eisenberg et al. 1975, as cited by Mudan 1984). According to Mudan (1984), exposure times for nuclear
explosions are typically very short and interpretation of the data is "somewhat subjectivV"
In a review of fire hazards by Takata in 1970 for the Armed Services Explosive Safety Board,
estimates are presented of critical radiant exposure necessary to ignite or damage several types of targets,
including people. As reported in Vol. I of the CPIA "Hazards of Rockets and Propellants," minimum
critical exposures necessary to ignite or damage people range from 94.6 kW/m2 for a 10-second exposure
time to 35.1 kW/m2 for a 60-second exposure and 27.8 kW/m2 for a 190-second exposure.
The WHAZAN manual includes a table for fatality based upon exposure to heat. Exhibit C-9
presents the lethality of heat flux as presented in the manual.
The exposure times given for the fatality levels are calculated from a Probit equation (indicating
statistical probability based on deviations from the mean of normal distribution), as follows:
Probit = -14.9 + 2.56 log,, (t x Q4/3 x 10"4)
where Q is in watts per square meter and t is in seconds
The WHAZAN model uses radiation levels of 12.5 to 37.5 kW/m> in its output, based on the
likelihood that people would be able to "shelter" from the fire within 30 seconds to one minute. For this
period of time, the probability of fatality ranges from 1 percent at 12.5 kW/m2 for 30 seconds to 99
percent at 37.5 for 50 seconds (see Exhibit C-9).
Mudan (1984) estimated a slightly lower fatality threshold of 10 kW/mJ for a 40-second exposure
time. Mudan estimated this level from the data of Eisenberg et al (1975, as cited by Mudan 1984). A plot
of thermal radiation versus time for injuries and fatalities shows that 1 percent fatalities may occur at 10
kW/mJ at a time of 40 seconds. The 10 kW/m3 radiation level suggested by Mudan, which is used in the
ARCHIE model to determine the zone for fatalities from pool fires, agrees quite well with the 12.5
kW/m» level.
Nazario (1988) indicates that all personnel should be evacuated from areas where the radiant heat
is likely to exceed one-half of the threshold value for second degree burns to bare skin (5,000 Btu/hr-ft2, or
16 kW/m2, for an exposure of about 5 seconds).
-------
-C-13-
EXHIBIT C-8
HUMAN INJURY CRITERIA
(Thermal Radiation Effects on Bare Skin)
Thermal
Radiation
1.75
6.4
10
12.5
37.5
Time of Exposure
(Seconds')
60
Injury or Effect
Pain threshold reached
Source
WHAZAN 1988
Buettner 1951
Hardy «t al. 1953
Stoll and Greene 1959
Bigelow et al. 1945
27
92
15
13
40
8
20
40
30
80
200
8
20
50
Severe pain
Second degree burn
Pain threshold reached
"Unbearable" pain
Second degree burn
(injury threshold)
Pain threshold reached
Second degree burn
Fatality threshold
1% fatality
50% fatality
99% fatality
1% fatality
50% fatality
99% fatality
HCHAP 1989
WHAZAN 1988
Mudan 1984
Mudan 1984
WHAZAN 1988
WHAZAN 1988
Mudan 1984
WHAZAN 1988
WHAZAN 1988
WHAZAN 1988
WHAZAN 1988
WHAZAN 1988
WHAZAN 1988
-------
-C-14-
Exhibit C-9
FATALITY FROM HEAT EXPOSURE
Thermal Radiation Seconds exposure for % fatality
(kW/m2) 1% 50% 99%
1.6
4.0
12.5
37.5
500
150
30
8
1300
370
80
20
3200
930
200
50
Reference: WHAZAN Handbook (1988)
-------
-C-15-
C.3.2 Distance Determination for Different Heat Radiation Levels
Modeling was carried out using the WHAZAN model to determine the effect of heat radiation
level on consequence distances for BLEVEs and pool fires.
C.3.2.1 BLEVES
For BLEVEs, ethylene and propylene were modeled at 10,000 pounds and 100,000 pounds. The
model was run for three heat flux consequence levels, 4 kW/m2, 12.5 kW/m2, and 37.5 kW/m2. The
relationship between the consequence distance results for ethylene and propylene and other hydrocarbon
results for BLEVEs can be seen by examining Exhibit C-10. The results curves for both quantities and
both hydrocarbons modeled are basically similar in shape but of differing magnitudes. The results curves
clearly are not linear: the change in magnitude between the consequence distance obtained at 4 kW/m2
and that at 12.5 kW/m2 is clearly greater than the change between 12.5 kW/m2 and 37.5 kW/m2 for all four
curves. Results for other hydrocarbons studied are approximately the same as the results for propylene.
C.3.2.2, Pool Fires
For pool fires, ethylene and propylene were modeled at 10,000 pounds and 100,000 pounds. The
WHAZAN model was run for three heat flux consequence levels, 1.6 kW/m2, 4 kW/m2, and 12.5 kW/m2.
Modeling results are presented in Exhibit C-ll. The results curves for both quantities and both chemicals
modeled are basically similar in shape but of only slightly differing magnitudes. The results curves clearly
are not linear: the change in magnitude between the consequence distance obtained at 1.6 kW/m2 and that
at 4 kW/m2 is clearly greater than the change between 4 kW/m2 and 12.5 kW/m2 for all four curves.
Results for other hydrocarbons studied are approximately the same as the results for ethylene and
propylene.
-------
Exhibit C-10
HAZARD CRITERIA COMPARISON FOR BLEVES,
INSTANTANEOUS RELEASES
CD
O
c
D
en
Q
800
700 -1
600-
500-
400-
300-
200-
100-
0
• Ethyiene (100.000 Iba.)
A Propylene (100.000 Iba.)
* Ethyiene (10.000 Iba.)
O Propylene (10.000 Iba.)
12.5
37.5
BLEVE Heat Flux (kW/m2)
-------
Exhibit C-ll
HAZARD CRITERIA COMPARISON FOR POOL FIRES,
INSTANTANEOUS RELEASES
600
0>
O
c
O
en
Q
500
400
300
• Ethyiene (100.000 Iba.)
A Propylene (100.000 Iba.)
# Ethyiene (10.000 Iba.)
O Propylene (10.000 Iba.)
^ 200
100
o
0
T"
1.6
12.5
15
Poolfire Heat Flux (kW/m2)
-------
-C-18-
C.4 References
AlChE. 1987. American Institute of Chemical Engineers. Methods for Calculation of Fire and Explosions
Hazards. New York: AIChE.
Brasie, W., Simpson, D. 1968. "Guidelines For Estimating Damage Explosion," AICHE Symposium on
Loss Prevention in the Process Industries, February 18-23, 1968.
CPLA. 1984. Hazards of Chemical Rockets and Propellants, Volume I, Safety, Health, and the
Environment. CPIA Publication 394, September 1984. Laurel, MD: Johns Hopkins University.
Eichler, T., Napadensky, H. 1977. "Accidental Vapor Phase Explosions on Transportation Routes Near
Nuclear Plants," IIT Research Institute, Prepared for Argonne National Laboratory. April 1977.
Eisenberg N.A. et al. 1975. Vulnerability Model. A Simulation System for Assessing Damage Resulting
from Marine Spills. NTIS AD-A015-245. Springfield, VA.
Fletcher E.R., Richmond D.R., Yelverton J.T. 1980 (May 30). "Glass Fragment Hazard from Windows
Broken by Airblast." Washington, D.C.:Defense Nuclear Agency. Report Number DNA 5593T.
Grelecki, C. "Fundamentals of Fire and Explosion Hazards Evaluation," C. Grelecki, Hazards Research
Corp., AICHE Today Series.
HCHAP. 1989. Handbook of Chemical Hazard Analysis Procedures. Federal Emergency Management
Agency, U.S. Department of Transportation, U.S. Environmental Protection Agency.
Kokinakis W., Rudolph R. 1982. An Assessment of the Current State-of-the-Art of Incapacitation by Air
Blast. Minutes of the 20th Explosives Safety Seminar, August 1982.
Lees, P.P. 1980. Loss Prevention in the Process Industries, Vol. 1. London: Butterworths.
McRae, T. 1984. "The Effects of Large Scale LNG/Water RPT Explosions," McRae, T.G., et.al., Lawrence
Livermore National Laboratory. April 27, 1984.
Mudan K.S. 1984. Thermal Radiation Hazards from Hydrocarbon Pool Fires. Prog. Energy. Combust.
Sci. 1984, Vol. 10. Great Britain: Pergamon Press Ltd.
Nazario N.N. 1988. Preventing or Surviving Explosions. Chemical Engineering, August 15, 1988.
NFPA. "Fire Protection Handbook," 16th edition, National Fire Protection Association (NFPA).
Richmond, D. 1968. "The Relationship Between Selected Blast-Wave Parameters and the Response of
Mammals Exposed to Air Blast." Richmond DR, Damon E.G., Retcher E.R., Bowen I.G., White C.S.,
Ann. N.Y. Acad. Sci., 1968.
Richmond D.R., Fletcher E.R. 1971. Blast Criteria for Personnel in Relation to Quantity-Distance.
Minutes of the 13th ASESB Seminar, 401-419.
U.S. Air Force 1983. "Explosives Safety Standards," AF Regulation 127-100, 20 May 1983.
U.S. DOT 1988. "Hazard Analysis of Commercial Space Transportation," Office of Commercial Space
Transportation Licensing Programs Division, U.S. Department of Transportation. May 1988.
WHAZAN. 1988. WHAZAN User Guide. Technica International Ltd.
-------
-C-19-
Wiekema, B. J. 1984. "Vapor Cloud Explosions - An Analysis Based on Accidents," Journal of Hazardous
Materials, 8 (1984).
-------
-D-l-
APPENDIX D
COMPARISON OF RESULTS OF MODELING BY DIFFERENT METHODS AND
RESULTS OF VARYING RELEASE SCENARIOS AND METEOROLOGICAL CONDITIONS
D.I Vapor Cloud Explosion Results
Exhibit D-l presents the results of modeling using several methods for vapor cloud explosions based
on 0.5 psi overpressure. This exhibit shows the maximum distance at which the models indicate that the
chosen overpressure would be reached. For releases of 100,000 pounds of gases, distances of 326 to 1,165
meters were calculated.
*
The distances obtained from WHAZAN, the Yellow Book, and ARCHIE at each overpressure level
agree quite well with each other, although the WHAZAN and Yellow Book distances for vapor cloud
explosions for ethylene and propane are significantly lower than the distances obtained from ARCHIE..
These two gases are assumed to be stored at reduced temperatures. The WHAZAN and Yellow Book
vapor cloud explosion calculations are based oh the quantity of gas immediately flashed into vapor on
release, and this quantity depends on the difference between the temperature of the released gas and its
boiling point. As ethylene and propane are assumed to be stored at temperatures not far above their
boiling points, the fraction calculated to be immediately flashed into vapor (i.e., the quantity that may be
involved in a vapor cloud explosion) is relatively small. ARCHIE, on the other hand, assumes that all the
material in a container of pressurized gas is released to the air as a gas-aerosol mixture unless the storage
temperature is more than 10°C below the boiling point of the gas. For ethylene and propane, the assumed
storage temperature, though below ambient temperature, is greater than the boiling point; therefore,
ARCHIE, unlike WHAZAN and the Yellow Book, does not yield consequence distances for these two
gases that are significantly lower than those for the other gases. If the storage temperature is more than
10°C below the boiling point, ARCHIE assumes the material is released as a liquid, and, therefore, no
vapor cloud explosion results are obtained.
ARCHIE produced vapor cloud explosion results for the liquids modeled; however, a disclaimer was
provided in the accompanying manual and on the output results pointing out that "incidents involving
tanks that are designed to operate at atmospheric pressure are very rare." An additional disclaimer stated
that clouds or plumes containing less than 1,000 pounds of vapor or gas are very unlikely to explode when
completely unconfined; modeling indicated that for heptane, toluene, and p-xylene, there would be less
than 1,000 pounds of vapor released. Because of these disclaimers, the ARCHIE vapor cloud explosion
results for the liquids modeled are not included in Exhibit D-l.
D.2 Vapor Cloud Fire Results
Exhibit D-l includes vapor cloud fire results calculated by five methods for releases of 100,000
pounds. The distances given are downwind distances for dispersion to a concentration equal to 50 percent
of the LFL. For the Yellow Book, results for gases were calculated based on both the quantity of gas
immediately flashed into vapor upon release and on the remaining liquefied gas evaporating from a pool
after flashing; note that the distances obtained using the quantity flashed into vapor are significantly larger
than those calculated using the evaporating pdol, except in the cases of the two gases assumed to be stored
at reduced temperatures. For these two gases, ethylene and propane, the distances calculated based on the
evaporating pool are greater than the distances based on the quantity flashed. Results from the AIChE
course material for gases are based only on the quantity flashed, as recommended. The distances
calculated for ethylene and propane, the gases assumed stored at reduced temperatures, are considerably
lower than the distances for the two gases assumed stored at ambient temperature. Distances obtained for
gases using the Green Book methodology are generally lower than those found by the other methods,
probably because the Green Book recommends assuming total release of the gas over a ten minute period,
while the other methods assume an instantaneous release. The distances calculated using the Green Book
method, which does not take storage temperature into account, are higher than other results only in the
cases of the Yellow Book results for the flashed fraction of ethylene and propane (the gases assumed
stored at reduced temperatures).
-------
-D-2-
Exhibit D-l
RESULTS OF MODELING BY ALL METHODS
I. WHAZAN Modeling Results for Instantaneous Releases, Moderate and Worst Case Meteorology, and
Prolonged Releases, Moderate Meteorology (100,000 pounds)
Chemical
Eihylenc
Propylene
Propane
Butane
Ethylenc
Oxide
Ethyl
Chloride
Acetaldehyde
Pentane
Acetone
Gasoline
Hexane
Trichloro-
ethylene
Heptane
Toluene
Acetic Acid
p-Xylene
Benzaldehyde
Dimethyl
Sulfoxide
Instantaneous Release
Moderate Meteorology
Distance (meters)
Boiling
Point
(K)
171
227
229
275
282
286
293
313
329
332
345
360
375
383
391
410
452
462
VCE
(0.5 psi)
457
1121
392
766
443
268
130
• *
• *
*»
»*
*•
• *
*•
*•
*»
**
**
VCF
(50%LFL)
1390
706
747
460
338
262
451
341
214
338
233
52
173
154
98
110
30
41
BLEVE
(12.5
kW/m2)
388
287
278
269
206
178
200
269
• •
• •
• *
• •
*•
• •
• *
*•
*•
*•
Pool
Fire
(12.5
kW/m2)
181
160
183
176
109
97
107
182
116
175
174
*•
168
153
**
152
109
86
Instantaneous Release
Worst Case Meteorology
Distance (meters)
VCE
(0.5 psi)
457
1121
392
766
*
*
*
• *
•
*•
• *
*
**
**
*
**
•
*
VCF
(50% LFL)
6350
2900
3100
1790
*
*
*
1280
*
1270
842
*
604
535
•
374
»
*
BLEVE
(12.5kW/mz)
388
287
278
269
*
•
* '
269
*
• *
• *
*
**
**
*
**
*
•
Pool
Fire
(12.5
kW/m2)
181
160
183
176
*
•
•
182
»
175
174
•
168
153
*
152
• •
»
Prolonged Release
Moderate Meteorology
Distance (meters)
VCF
(50%LFL)
79
18
73
20
*
*
*
21
•
64
56
*
57
67
•
65
*
»
BLEVE
(12.5kW/m2)
388
287
278
269
*
»
*
269
•
• •
*•
•
**
*#
•
•»
• »
*
*
Pool
Fire
(12.5kW/m2)
17
15
17
9
•
*
*
23
*
16
15
•
15
16
*
16
*
*
Jet
Fire
(12.5kW/m2)
•*
42
**
21
*
»
*
*#
*
**
«*
•
**
**
*
»#
*
*
* Analysis was not carried out
** No results were obtained from analysis, or results were not included, based on evaluation.
VCE = Vapor Cloud Explosion
VCF = Vapor Cloud Fire
-------
-D-3-
Exhibit D-l (continued)
RESULTS OF MODELING BY ALL METHODS
II. Modeling Results Using Other Methods for Instantaneous Releases, Moderate Meteorology
(100,000 pounds)
Chemical
Ethylene
Propylene
Propane
Butane
Pentane
Gasoline
Hexane
Heptane
Toluene
p-Xylene
ARCHIE Results
Distance (meters)
Boiling
Point
(K)
171
227
229
275
313
332
345
375
383
410
VCE
(0.5 psi)
867
682
683
680
• *
*•
• ft
• •
• *
**
VCF
(50% LFL)
770
751
733
712
453
447
160
112
97
0
BLEVE
(10kW/m2)
285
285
285
285
ft*
ft*
ft*
ft*
»•
• *
Pool
Fire
(10kW/m2)
*•
*•
• •
• *
151
137
151
153
150
144
Yellow Book Results
Distance (meters)
VCE
(0.5 psi)
415
932
326
635
• •
• »
**
• •
• •
• »
VCF (Gases)
Flash
Fraction
(50% LFL)
330
862
261
547
ft*
ft*
• ft
ft*
. *•
ft*
VCF (All)
Pool
Evaporation
(50% LFL)
922
554
586
242
239
216
115
60
42
22
Pool
Fire
(12.5
kW/m2)
134
112
137
120
120
107
112
106
86
86
AICbE Manual Results
Distance (meters)
VCF
(Gases)
Flash
Fraction
(50% LFL)
452
1165
355
720
• •
• •
*•
• •
*»
»•
VCF
(Liquids)
Pool
Evaporation
(50% LFL)
**
• *
• *
**
187
170
95
53
38
21
Pool
Fire
(12.5kW/m2)
299
188
225
190
221
213
206
202
169
166
Green Book
Results
Distance
(meters)
VCF
(50% LFL)
371
350
332
310
165
163
86
50
44
28
+
+
+
+
* Analysis was not carried out.
••No results were obtained from analysis, or results were not included, based on evaluation.
+ Method is not valid for distances less than 100 meters. .
VCE = Vapor Cloud Explosion
VCF = Vapor Cloud Fire
-------
-D-4-
Results for chemicals that are liquids at ambient temperatures are based on evaporating pools for
all the models (see Appendix B for details). Distances to a concentration equal to 50 percent of the LFL
for vapor cloud fires range from 261 to 1,390 meters for gases; distances are generally smaller for liquids,
although significant distances result for volatile liquids (e.g., distances calculated by the various methods
range from 165 to 453 meters for pentane). Distances obtained using the Green Book dispersion modeling
calculations are not valid for less than 100 meters; therefore, the Green Book results for four relatively
non-volatile liquids should be disregarded. It is possible that distances less than 100 meters obtained from
some of the other methods may also be invalid, as similar methods appear to be used in the calculations.
D.3 BLEVE Results
»
Exhibit D-l presents the results obtained using WHAZAN and ARCHIE to model BLEVEs for
100,000 pound releases. WHAZAN results are shown for a heat radiation level of 12.5 kW/m2 for
WHAZAN. The distance reported as the "fatality distance" on the ARCHIE output results is presented in
this exhibit. BLEVEs are not modeled in ARCHIE; however, since fireballs are generally associated with
BLEVEs, these results from the ARCHIE model are presented in Exhibit D-l for comparison to
WHAZAN BLEVE results. The other models used for consequence analysis did not include methods for
analyzing BLEVEs.
ARCHIE results are based on fireball size and heat radiation (see Appendix C, Section A. 1.10).
According to the ARCHIE manual, the fatality level corresponds to a heat flux in excess of 160 kilojoules
per square meter (kJ/m2). Using the fireball duration of 15 seconds to convert the heat radiation level
from kJ/m2 to kW/m2 indicates that the ARCHIE fatality level is about 11 kW/m2, which is very close to
12.5 kW/m2, used by WHAZAN. Note, however, that WHAZAN gives a BLEVE duration of 36 seconds
while ARCHIE gives a duration of 15 seconds for a 100,000 pound release.
The ARCHIE model gives a fatality distance of 285 meters for all the chemicals modeled; the lack
of variation by chemical is probably due to the fact that, according to the ARCHIE manual, ARCHIE
assumes that the chemical modeled is propane or has similar characteristics.. WHAZAN consequence
distances range from 178 meters to 387 meters for a heat radiation level of 12.5 kW/m2.
D.4 Pool Fire Results
Results of modeling using different methods for pool fires, showing the maximum distance at which
a heat radiation level of 12.5 kW/m2 (10 kW/m2 for ARCHIE) is reached, are presented in Exhibit D-l.
The WHAZAN model indicated that the pool fires of the chemicals modeled would last 60 seconds or
more; as stated in the WHAZAN manual, it might be possible to "shelter" from the fire within about 30
seconds to one minute.
Differences between chemicals modeled by each method are relatively small. The distances
calculated by different methods are in reasonably good agreement with each other. The calculations based
on the AIChE-sponsored materials gave the greatest distances (166 to 299 meters), while the Yellow Book
method gave the smallest (86 to 134 meters). The ARCHIE model does not give pool fire results for
gases. The other models allow the calculation of results for gases; the results for gases differ little from
the results for liquids. The distances found for pool fires for 100,000 pound releases of gases for a heat
radiation level of 12.5 kW/m2 are smaller than the distances for vapor cloud explosions for the same
quantity modeled at 3.0 psi or lower, except for gases assumed stored at reduced temperatures. Vapor
cloud fire results (for 50 percent of the LFL) and BLEVE results (for 12.5 kW/m2) for gases are larger
than the pool fire results. For relatively volatile liquids (e.g., pentane), vapor cloud fire distances are
generally greater than pool fire distances; for less volatile liquids, distances calculated by most of the
methods are greater for pool fires.
-------
-D-5-
D.5 Jet Fire Results
Jet fire results, which are, by definition, prolonged releases, were obtained only for propylene and
butane, the two gases assumed to be liquefied under pressure. For jet fires, the release was assumed to
take place from a 1.5 inch hole in a tank 12 feet in diameter. Results are shown for 100,000 pound
releases in Exhibit D-l for a heat flux level of 12.5 kW/m2. Jet fires of propylene and butane under these
conditions gave distances that are smaller than any of the distances for any other accident type resulting
from instantaneous releases. Distances for jet fires are somewhat greater, however, than for vapor cloud
fires or pool fires resulting from prolonged releases under the same conditions.
D.6 Instantaneous Release Compared to Prolonged Release »
Ten hydrocarbons studied were modeled using the prolonged release scenario in the WHAZAN
linked model for comparison with the modeling results for instantaneous release. The WHAZAN linked
model (prolonged release) was run for 100,000 pounds of each chemical assuming the same meteorological
conditions (3.0 meters per second wind speed and D wind stability), and storage conditions as for the
instantaneous release. To simulate a prolonged release, a 1.5 inch hole in a 12 foot diameter tank was
assumed. The average evaporation rate for the pool, as determined in the linked model, was then used to
run the WHAZAN stand-alone model for buoyant plume dispersion (see Appendix B.I for more
information).
Exhibit D-l shows results of WHAZAN modeling for prolonged releases of ten chemicals, where
release was assumed to take place from 1.5 inch diameter holes in 12 foot diameter tanks. Distances to 50
percent of the LFL for vapor cloud fires are much smaller (18 to 79 meters) than for instantaneous
releases.
For pool fires, prolonged releases also result in much smaller consequence distances than
instantaneous releases. As was the case for instantaneous releases, the consequence distance varies little
from chemical to chemical. The magnitude of the results curve is much smaller for prolonged releases
than for instantaneous releases, and all of the prolonged release consequence distances are less than 25
meters.
WHAZAN did not produce vapor cloud explosion results for prolonged releases. As shown in
Exhibit D-l, WHAZAN produced BLEVE results for prolonged releases that are identical to the results
for instantaneous releases. It is not clear what these results mean, since BLEVEs seem to be
instantaneous releases by definition (see Section 6.3).
D.7 Moderate Meteorology Compared to Worst Case Meteorology
Ten hydrocarbons were modeled using WHAZAN to determine the effect on consequence distances
of using worst case meteorology rather than moderate meteorology. The WHAZAN linked model
(instantaneous release) was run for 100,000 pounds of each chemical using the same storage temperatures
and pressures as used with moderate meteorology. Worst case meteorological conditions (1.5 meters per
second wind speed and F wind stability) were used rather than moderate meteorological conditions (3.0
meters per second wind speed and D wind stability). The average evaporation rate for the pool, as
determined in the linked model, was used to run the WHAZAN stand-alone model for buoyant plume
dispersion. Various other default values were used in the model; these values are summarized in Appendix
E.
-------
-D-6-
Exhibit D-l includes vapor cloud fire results from WHAZAN for ten chemicals using worst case
meteorological conditions of wind speed of 1.5 meters per second and F atmospheric stability. The
WHAZAN results appear to be strongly dependent on the meteorological assumptions made, with
distances determined using worst case meteorology being much greater than those resulting from modeling
using moderate meteorology. As shown in Exhibit D-l, distances under worst case meteorological
conditions extend from 1,790 to 6,350 meters for gases, compared to 460 to 1,390 meters under moderate
conditions. For the other consequence types, the results were the same for both meteorological
conditions, indicating that varying meteorological conditions probably have little effect on the potential
consequences of vapor cloud explosions, BLEVEs, and pool fires.
-------
-E-l-
APPENDIX E
INPUTS FOR MODELING
This appendix presents the inputs for the consequence analyses. Temperatures and pressures
assumed as storage conditions for each of the chemicals modeled are listed in Exhibit E-l. Exhibit E-2
presents the physical/chemical property data used for modeling. The data shown were used for WHAZAN
modeling, which required more data than the other methods. The same data were used as required for
modeling by the other methods; in some cases the methods required conversion of the data to different
units. Exhibit E-3 shows the assumptions used for WHAZAN modeling for instantaneous releases and
moderate meteorology. The same meteorological assumptions were used in modeling by the other
methods. Assumptions for WHAZAN modeling of prolonged releases are shown in Exhibit E-4, and
assumptions for WHAZAN modeling under worst-case meteorological conditions are shown in Exhibit E-
5. Modeling for prolonged releases and for worst-case meteorology was done only with WHAZAN.
Exhibit E-6 presents the data and assumptions used for ARCHIE modeling.
-------
-E-2-
Exhibit E-l
Temperatures and Pressures Used in Scenarios
Storage Conditions
Chemical
Gases
Ethylene
Propylene
Propane
Butane
Ethylene oxide
Ethyl chloride
TemD.no
175
293
232
293
293
288
Press.CBars)
1.3
13.3
1.3
6.3
6.18
1.3
Acetaldehyde " 293
Pentane 293
Gasoline 293
Hexane 293
Heptane 293
Toluene 293
p-Xylene 293
Acetone 293
Trichloroethylene 293
Acetic acid 293
Benzaldehyde 293
Dimethyl sulfoxide 293
1.013
2.74
1.013
1.013
1.013
1.013
1.013
1.013
1.013
1.013
1.013
1.013
-------
-E-3-
Exhibit E-2
Chemical Data Used for Models
A. Hydrocarbons
PROPERTIES
CAS Number
Liquid Heat Capacity (J/kg/K)
Liquid Density (kg/cube m)
Gamma Ratio of Specific Heat
Heat of Evaporation (J/kg)
Heat of Combustion (J/kg)
Lower Flammability Limit (tract.)
Upper Flammability Limit (tract.)
Boiling Point at 1 Atmos. (K)
Molecular Weight
Vapor Heat Capacity (J/kg/K)
Critical Temperature (K)
Saturated Vapor Pressure (Bara)
Liquid Enthalpy (J/kg)
Vapor Enthalpy (J/kg)
Vapor Entropy (J/kg/K)
Saturated Vapor Temperature
(K)
Vapor Density (kg/cube m)
Ethylene
74851
2750.0
569.0
1.255
4.83e+05
-4.72e+07
0.027
0.286
170.828
28.05
700.0
283.1
--
-6.69e+05
-396000
-131.1
171.049
1.163
Propylene
115071
2500.0
520.0
1.152
4.38e+05
-4.58e+07
0.020
0.111
227.423
42.08
1000.0
365.0
11.717
-4.35e+05
-65000
-131.7
227.71
1.744
Propane
74986
2400.0
590.0
1.13
4.26e+05
-4.60e+07
0.021
0.095
229.273
44.09
1100.0
369.8
9.622
•4.33C+05
-62000
-J44.2
229.575
1.828
Butane
75285
2500.0
600.0
1.092
3.90e+05
-4.54e+07
0.018
0.084
274.843
58.12
1200.0
425.0
2.282
-4.37C+05
-81000
-54.4
275.183
2.409
Penlane
109660
2207.0
626.0
1.086
3.67e+05
-4.54C+07
0.015
0.078
312.669
72.15
1547.0
469.7
0.593
-4.33C+05
-40400
-157.0
313.056
-
Gasoline
8006619
2056.0
•t
732.0
1.054
3.20e+05
-4.35C+07
0.014
0.074
332.226
72.00
1234.0
787.0
0.501
0.00
390000
-1429.0
332.952
-
Hexane
110543
2456.19
659.0
1.063
3.35e+05
-4.48e+07
0.014
0.074
344.589
86.17
1819.0
507.4
0.175
-4.25e+OS
-40200
-171.6
344.995
„
Heptane
142825
2605.74
683.8
1.045
3.17e+OS
-4.46e+07
0.012
0.067
374.557
100.21
1992.26
540.7
0.055
-4.23e+05
-40200
-203.3
374.993
-
Toluene
108883
1738.0
867.0
1.089
3.61 e+ 05
-4.06e+07
0.012
0.070
333.111
92.00
1122.0
592.0
0.039
3.17e+05
746000
-2732.6
383.553
-
p-Xylene
106423
1750.0
861.0
1.071
3.40e+05
-4.08e+07
0.011
0.070
410.05
106.00
1193.0
616.2
0.012
0.00
417000
-1527.5
410.509
-
-------
-E-4-
Exhibit E-2
Chemical Data Used Tor Models
B. Non-Hydrocarbons
PROPERTIES
CAS Number
Liquid Heat Capacity (J/kg/K)
Liquid Density (kg/cube m)
Gamma Ratio of Specific Heat
Heat of Evaporation (J/kg)
Heat of Combustion (J/kg)
Lower Flamraability Limit (fract.)
Upper Flammabilily Limit (fract.)
Boiling Point at 1 Atmos. (K)
Molecular Weight
Vapor Heat Capacity (J/kg/K)
Critical Temperature (K)
Saturated Vapor Pressure (Bara)
Liquid Enthalpy (J/kg)
Vapor Enthalpy (J/kg)
Vapor Entropy (J/kg/K)
Saturated Vapor Temperature (K)
Vapor Density (kg/cube m)
Acetaldehyde
75070
2483.09
783.4
1.182
5.69e+05
-2.51e+07
0.016
0.104
293.55
44.053
946.63
461.0
1.198
7.40e+05
282237
5679.52
298.15
--
Acetone
67641
2176.0
789.9
1.127
5.11e+05
-2.86e+07
0.026
0.128
329.44
58.08
1296.7
0.306
513671
2.84e+05
5085.23
298.15
--
--
Dimethyl
Sulfoxide
67685
1957.63
1095.4
--
6.03e+05
-1.98e+07
0.026
0.285
462.15
78.129
459.88
726.0
0.000812
S.84e+05
i 137113
3920.06
298.15
--
Trlchloro-
Elhylene
79016
946.58
1464.2
1.116
2.40e+04
-6.58e+06
0.125
0.90
360.1
131.389
390.38
571.0
0.0983
2.82e+05
116392
2474.73
298.15
--
Ethyl
Chloride
75003
1616.98
897.8
1.155
3.79e+05
-1.99e+07
0.038
0.154
285.42
64.514
970.9
460.35
1.588
3.42e+05
205523
4274.73
298.15
2.87
Ethylene
Oxide
75218
1979.0
869.0
1.212
5.80e+05
-2.67C+07
0.03
1.00
281.814
44.05
1006.0
469.0
1.699
-6.29e+05
-26400
-100.8
282.118
1.82*6
Acetic
Acid
64197
2048.2
1042.9
1.145
4.05e+05
-1.31e+07
0.054
0.16
391.05
60.052
1114.0
592.71
0.016
4.72e+05
226460
4197.53
294.30
-
Benzaldehyde
100527
1676.3
1041.5
1.10
3.62e+05
-3.20e+07
0.015
0.10
452.2
106.124
981.8
694.8
0.0009
5.00e+05
292724
--
297.00
-
-------
-E-5-
Exhibit E-3
WllAZAN Modeling Data for Instantaneous Releases, Moderate Meteorology
A. Hydrocarbons
Data Inputs
Storage Temperature (K)
Storage Pressure (Bar)
Ambient Temperature (K)
Stored Mass (kg)
Bund Area (m2)
Wind Speed (m/s)
Max Time of Interest (s)
BUOYANT PLUME DISPERSION
Effective Release Height (m)
Release Rate (kg/s)
Min. Cone, of Interest (ppm)
t
Wind Speed (m/s)
Ambient Temperature (K)
Surface Roughness Parameter
Atmospheric Stability Category
Ethylene
175.0
1.3
293.0
45360
0.000
3.000
100000
0.000
761.0
13500
3.000
293.0
000.1
D
Propylene
293.0
13.3
293.0
45360
0.000
3.000
100000
0.000
265.7
10000
3.000
293.0
000.1
D
Propane
232.0
1.3
293.0
45360
0.000
3.00
100000
0.000
244.5
10500
3.00
293.0
000.1
D
Butane
293.0
6.3
293.0
45360
0.000
3.000
100000
0.000
157.4
9000
3.000
293.0
000.1
D
Pentane
293.0
2.74
293.0
45360
0.000
3.000
100000
0.000
96.25
7500
3.000
293.0
000.1
D
Gasoline
293.0
1.013
293.0
45360
0.000
3.000
100000
0.000
88.25
7000
3.000
293.0
000.1
D
Hexane
293.0
1.013
293.0
45360
0.000
3.000
100000
0.000
54.81
7000
3.000
293.0
000.1
D
Heptane
293.0
1.013
293.0
45360
0.000
3.000
100000
0.000
31.94
6000
3.000
293.0
000.1
D
Toluene
293.0
1.013
293.0
45360
0.000
3.0
100000
0.000
24.04
6000
3.000
293.0
000.1
D
p-Xykne
293.0
1.013
293.0
45360
0.000
3.0
100000
0.000
14.11
5500
3.000
293X>
000.1
D
-------
-E-6-
Exhibit E-3
WI1AZAN Modeling Data for Instantaneous Releases, Moderate Meteorology
B. Non-Hydrocarbons
Data Inputs
Storage Temperature (K)
Storage Pressure (Bar)
Ambient Temperature (1C)
Stored Mass (kg)
Bund Area (m2)
Wind Speed (m/s)
Max Time of Interest (s)
BUOYANT PLUME DISPERSION
Effective Release Height (m)
Release Rate (kg/s)
Min. Cone, of Interest (ppm)
Wind -Speed (m/s)
Ambient Temperature (K)
Surface Roughness Parameter
Atmospheric Stability Category
Ethylene
Oxide
293.0
6.18
293.0
45360
0.000
3.000
100000
0.000
109.6
15000
3.000
293.0
0.100
D
Elhyl
Chloride
288.0
1.3
293.0
45360
0.000
3.000
100000
0.000
137.4
19000
3.000
293.0
0.100
D
Acetaldehyde
293.0
1.013
293.0
45360
0.000
3.000
100000
0.000
102.5
8000
3.000
293.0
0.100
D
Acetone
293.0
1.013
293.0
45360
0.000
3.000
100000
0.000
44.51
13000
3.000
293.0
0.100
D
Trlchloro
ethylene
293.0
1.013
293.0
45360
0.000
3.000
100000
0.000
44.52
62500
3.000
293.0
0.100
D
Acetic Acid
293.0
1.013
293.0
45360
0.000
3.000
100000
0.000
30.84
27000
3.000
293.0
0.100
D
Benzaldehyde
293.0
1.013
293.0
45360
0.000
3.000
100000
0.000
1.38
7500
3.000
293.0
0.100
D
Dimethyl
Sulfoxlde
293.0
1.013
293.0
45360
0.000
3.000
100000
0.000
3.357
13000
3.000
293.0
0.100
D
-------
-E-7-
Exhibit E-4
WIIAZAN Modeling Data for Prolonged Releases
Data Inputs
LINKED MODEL
Storage Temperature (K)
Storage Pressure (Bar)
Ambient Temperature (K)
Stored Mass (kg)
Bund Area (m2)
Wind Speed (m/s)
Max Time of Interest (s)
•
BUOYANT PLUME DISPERSION
Effective Release Height (m)
Release Rate (kg/s)
Min. Cone, of Interest (ppm)
Wind Speed (m/s)
Ambient Temperature (K)
Surface Roughness Parameter
Atmospheric Stability Category
Ethylene
175.0
1.3
293.0
45360
0.000
3.000
100000
0.000
761.0
13500
3.000
293.0
000.1
D
Propylene
293.0
13.3
293.0
45360
0.000
3.000
100000
0.000
265.7
10000
3.000
293.0
000.1
D
Propane
232.0
1.3
293.0
45360
0.000
3.00
100000
0.000
244.5
10500
3.00
293.0
000.1
D :
Butane
293.0
6.3
293.0
45360
0.000
3.000
100000
0.000
157.4
9000
3.000
293.0
000.1
D
Pentane
293.0
2.74
293.0
45360
0.000
3.000
100000
0.000
96.25
7500
3.000
293.0
000.1
D
Gasoline
293.0
1.013
293.0
45360
0.000
3.000
100000
0.000
88.25
7000
3.000
293.0
000.1
D
Hexane
293.0
1.013
293.0
45360
0.000
3.000
100000
0.000
54.81
7000
3.000
293.0
000.1
D
Heptane
293.0
1.013
293.0
45360
0.000
3.000
100000
0.000
31.94
6000
3.000
293.0
000.1
D
Toluene
293.0
1.013
293.0
45360
0.000
3.0
100000
0.000
24.04
6000
3.000
293.0
000.1
D
p-Xylene
293.0
1.013
293.0
45360
0.000
3.0
100000
0.000
14.11
5500
3.000
293.0
000.1
D
-------
-E-8-
Exhibit E-5
WIIAZAN Modeling Data for Instantaneous Releases, Worst Case Meteorology
Data Inputs
Storage Temperature (K)
Storage Pressure (Bar)
Ambient Temperature (K)
Stored Mass (kg)
Bund Area (ra2)
Wind Speed (m/s)
Max Time of Interest (s)
BUOYANT PLUME DISPERSION
Effective Release Height (m)
Release Rate (kg/s)
Min. Cone, of Interest (ppm)
Wind Speed (m/s)
Ambient Temperature (K)
Surface Roughness Parameter
Atmospheric Stability Category
Ethylene
175.0
1.3
293.0
45360
0.000
1.500
100000
0.000
576.7
13500
1.500
293.0
0.100
F
Propylene
293.0
13.3
293.0
45360
0.000
1.500
100000
0.000
201.4
10000
1.500
293.0
0.100
F
Propane
232.0
1.3
293.0
45360
0.000
1.500
100000
0.000
244.5
10500
1.500
293.0
0.100
F :
Butane
293.0
6.3
293.0
45360
0.000
1.500
100000
0.000
119.3
9000
1.500
293.0
0.100
F
Pentane
293.0
2.74
293.0
45360
0.000
1.500
100000
0.000
72.95
7500
1.500
293.0
0.100
F
Gasoline
293.0
1.013
293.0
45360
0.000
1.500
100000
0.000
66.88
7000
1.500
293.0
0.100
F
Hexane
293.0
1.013
293.0
45360
0.000
1.500
100000
0.000
41.54
7000
1,500
293.0
0.100
F
Heptane
293.0
1.013
293.0
45360
0.000
1.500
100000
0.000
41.54
6000
1.500
293.0
0.100
F
Toluene
293.0
1.013
293.0
45360
0.000
1.500
100000
0.000
18.22
6000
1.500
293.0
0.100
F
p-Xylene
293.0
1.013
293.0
45360
0.000
1.500
100000
0.000
10.70
5500
1.500
293.0
0.100
F
-------
-E-9-
Exhibit E-6
ARCHIE Modeling Data
CAS Number
PHYSIOCHEMICAL PROPERTIES OF MATERIAL
Normal Boiling Point (deg F)
Molecular Weight
Liquid Specific Gravity
Vapor Pressure at Container Temp (psia)
Vapor Pressure at Ambient Temperature (psia)
Lower Flammable Limit (vol%)
Lower Heat of Combustion (Btu/lb)
Gas Explosion Yield Factor
CONTAINER CHARACTERISTICS
Total Weight of Contents (Ibs)
Temperature of Container Contents (deg F)
Tank Contents During Fireball (Ibs)
ENVIRONMENTAL/LOCATION
CHARACTERISTICS
Ambient Temperature (deg F)
Wind Velocity (mph)
Efhylene
74851
-151.91
28.05
-
18.90
--
2.7
20292
0.06
100000
-144.0
100000
68.0
11.9
Propylene
115071
-50.039
42.08
-
192.90
-
2.0
19690
0.03
100000
68.0
100000
68.0
11.9
Propane
74986
-46.71
44.09
-
18.85
-
2.1
19776
0.03
100000
-42.0
100000
68.0
11.9
Butane
75285
35.32
58.12
-
91.34
-
1.8
19519
0.03
100000
68.0
100000
. 68.0
11.9
Pentane
109660
103.404
72.15
0.626
7.12
7.12
1.5
19519
0.03
100000
68.0
100000
68.0
11.9
Gasoline
8006619
138.607
72.00
0.732
6.49
6.49
1.4
18702
0.03
100000
68.0
100000
68.0
11.9
Hexane
110543
160.86
86.17
0.659
2.05
2.05
1.4
19261
0.03
100000
68.0
100000
68.0
„ 11.9
Heptane
142825
214.803
100.21
0.6838
0.62
0.62
1.2
19175
0.03
100000
68.0
-
68.0
11.9
Toluene
108883
230.2
92.00
0.867
0.44
0.44
1.2
17455
0.03
100000
68.0
--
68.0
11.9
p-Xylene
106423
278.69
106.0
0.861
0.13
0.13
1.1
17541
0.03
100000
68.0
-
68.0
11.9
-------
F-l-
APPENDIX F
ADDITIONAL INFORMATION FROM ACCIDENT DATABASES
F.I Additional Information from the Acute Hazardous Events Data Base
The Acute Hazardous Events Data Base (AHE/DB) includes information on 773 incidents for
which one of the end effects (up to four end effects may be listed) was fire or explosion. Most of the
explosion incidents involved flammable chemicals rather than explosives and, therefore, were included in
this analysis of flammable chemical accidents. The first (or primary) end effect was listed a! fire or
explosion in 542 (70 percent) of such incidents; for most of the others, spill or vapor release was listed as
the first end effect, and fire or explosion was second. Of the 773 fire or explosion incidents, 652 (84
percent) took place at a fixed location, while 121 (16 percent) occurred during transportation.
A total of 431 deaths was reported resulting from 88 of the fire or explosion incidents, and 4,195
injuries were reported from 286 incidents. Fire and explosion incidents account for about 13 percent of
the total incidents in the AHE/DB, but they account for nearly 92 percent of the deaths ami about 36
percent of the injuries reported. Exhibit F-l graphically presents fire and explosion events and deaths and
injuries in fire and explosion events compared to the total in the AHE/DB.
Specific flammable chemicals that the AHE/DB reports as having been involved in two or more in-
plant fire or explosion incidents are presented in Exhibit F-2. Number of incidents versus boiling point (as
an indication of volatility) is shown; the DOT hazard class and packing group that would be assigned based
on flash point and boiling point is also indicated for each chemical (see key). In the case of some fuel
chemicals (gasoline, propane, butane, natural gas/methane) the number of in-plant incidents was estimated
from the number reported in the 1985 version of the AHE/DB, as fuels have been deleted from the 1988
version, by assuming that the number of incidents would have increased by a factor of 2. The total number
of incidents in the 1988 version of the data base was approximately double the number in the 1985 version;
therefore, it was assumed that the number of fuel incidents probably also would have approximately
doubled.
As Exhibit F-2 shows, a number of flammable gases were reported involved in in-plant fires or
explosions, as were several liquids in DOT Packing Group I/NFPA Flammability Rating 4 (i.e., the most
flammable and volatile categories). There were also a number of flammable chemicals that fall into DOT
packing groups II and III and several chemicals that are considered merely combustible rather than
flammable by DOT and NFPA standards. There does not appear to be any clear correlation between
physical/chemical properties (flash point and boiling point) and the number of fire and explosion incidents
reported for chemicals in the AHE/DB. The severity of the incidents, as indicated by the number of
deaths and injuries that resulted, was also considered in relation to the flash points and boiling points of
the chemicals. There did not appear to be a clear correlation between the properties of the chemicals and
the severity of the incidents. Accidents and resulting consequences are dependent not only upon the
physical/chemical properties of the chemicals involved, but also on factors such as location, time, weather,
amount, other chemicals nearby, and release conditions; little information on these factors is provided in
the AHE/DB.
Production volume for the flammable and combustible chemicals shown in Exhibit F-2 was
considered as a possible factor related to the number of incidents reported. Although several high-volume
chemicals were involved in relatively large numbers of in-plant incidents (toluene in 17 incidents, gasoline
in an estimated 13 incidents, vinyl chloride in 7 incidents), there appears to be no clear correlation
between number of incidents and production volume.
-------
F-2-
In addition to the specific chemicals shown in Exhibit F-2, there were many unknown chemicals and
non-specific materials involved in incidents reported in the AHE/DB. A large number of incidents (61)
involved unknown substances. Chemical types or categories listed include the following:
Alcohol: 3 '
Explosives: 14
Fertilizer: 4
Flammable chemicals, gas, liquid, or material: 7
Hazardous waste or industrial waste: 12
Paint: 7
Pesticide: 29 *
Petroleum products or petroleum refining slurry: 6
Plastics: 7
Rubber, rubber hose or tires: 8
Solvents: 13
Various other non-specific listings also are included, such as several kinds of oil and fuel, and non-
chemical materials such as insulated wire.
A number of chemicals reported to have been involved in fires or explosions are not flammable or
are only slightly flammable (e.g., chlorine (23 incidents), ammonia (19 incidents), hydrochloric acid (12
incidents), nitric acid (11 incidents)). These chemicals were not included in Exhibit 2; their involvement in
fires and explosions may be related to their reactivities and other chemicals involved in the incidents rather
than to flammability.
-------
-F-3-
EXHIBIT F-l
FIRE AND EXPLOSION EVENTS COMPARED TO ALL
EVENTS IN ACUTE HAZARDOUS EVENTS DATA BASE
Fire and Explosion Events Compared to Total Events
Deaths in Rre and Explosion Events Compared to Total Deaths
Injuries in Rre and Explosion Events Compared to Total Injuries
-------
Exhibit F-2
FLAMMABLE GASES AND LIQUIDS
REPORTED IN IN-PLANT RRES OR EXPLOSIONS
JJ
c
u
2
*o
c
—
c
0
1
a
ti
^
C
^_
o
i.
u
A
E
3
4U —
19 -
IB -
17 -
16 -
15-
14 -
13 -
12 -
11 -
10 -
9 -
8 -
7 -
6 -
5 -
4 -
3 -
2 -
1 -
lOtUM
O
MUUK
0
fWVMK
P
mp"U
,
vim M.OUK
P
•ff T
IfNRMMIK ' WW« "«*
HrMMTB P •• O O
fTmm sariK ISWIIMC faSi "lSiJ{m UMMM CMMH
P P P MMUCIMIH^ O O A
•IMMK HUM KtmUK IIM IMI •>.«.« MMIM
P P P 4-J" O «t»< O O mm
PHtsraiH HIMIIK t*mm* mmaMM KIIUUI W«MKMI«
D a O» ^ A A >t X X
KSNM »»«*
1 ii i i
-300 -100 O 100 30
_.„_-.«. _:
Boiling Point (degrees C)
• »lill ttlUIl
k tllCW.«MKIHMC
c KMIK
4 IIICMMOtlHUfM
-------
F-5-
KEY FOR EXHIBIT F-2
D Flammable gases (Boiling point <20°C)
DOT Hazard Class 2.1 (2.3 for flammable and poison)
NFPA Flammability Rating 4
+ Flammable liquids, DOT Packing Group I or equivalent
DOT Hazard Class 3, Packing Group I: Boiling point <35°C
NFPA Flammability Rating 4: Boiling point <38°C, Flash point <23°C
0 Flammable liquids, DOT Packing Group II or equivalent ;
DOT Hazard Class 3, Packing Group II: Boiling point >35°C, Flash point <23°C
NFPA Flammability Rating 3: Boiling point .>380C, Flash point <23°C,
or Flash point <38°C and Flash point _>23°C
A Flammable liquids, DOT Packing Group III or equivalent
DOT Hazard Class 3, Packing Group III: Boiling point >35°C, Flash point >230C,
Flash point <60.5°C
NFPA Flammability Rating 2: Flash point >38°C, Flash point .<93°C
X Combustible liquids
No applicable DOT Hazard Class
NFPA Flammability Rating 2 or 1 (NFPA 1: Flash point >93°C)
-------
F-6-
F.2 Additional Information from the ARIP, OSHA, M & M Accident Databases
This section provides additional accident information from the ARIP, OSHA, and M & M Databases.
Exhibit F-3 presents the flammable chemicals most frequently involved in fires and explosions. In the
ARIP and OSHA databases, the top five chemicals were listed. For the M & M database, the top 10 chemicals
were listed. The data indicate that simple hydrocarbon chemicals are most often involved in fires or explosions.
Specifically, the hydrocarbon chemicals most frequently listed are propane, propylene, butane, and methane.
Ethylene oxide also was involved in a number of incidents.
EXHIBIT F-3 ,
FLAMMABLE CHEMICALS INVOLVED IN FRIES AND EXPLOSIONS
Chemical
Acetylene
Butane
Cumene
Ethylene
Ethylene Oxide
Hydrogen
Methane/Natural Gas
Methyl Chloride
Propane
Propylene
Propylene Oxide
Silane
Styrene
Unspecified flammable
gas or liquid
Unspecified
hydrocarbon
Incidents
in ARIP
database
3
1
1
1
Incidents
in OSHA
database
2
2
5
8
9
Incidents
inM&M
database
4
2
2
3
3
15
6
2
2
4
Total
Incidents
2
6
2
2
6
3
5
1
23
7
2
1
2
9
4
Exhibit F-4 lists the ARIP accidents involving flammable chemicals. Although ARIP collects data on
deaths and injuries to the public, none of the flammable chemical fires or explosions involved public health impact.
The ARIP accidents indicate that equipment failure is a prevalent cause of flammable chemical explosions.
-------
-F-7-
EXHIBIT F-4
SUMMARY OF EXPLOSION ACCIDENTS INVOLVING FLAMMABLE CHEMICALS IN THE ARIP DATABASE
Chemical (s)
Ethylene Oxide
Ethylene Oxide
Ethylene Oxide
Methyl Chloride
Quantity
Released in
Pounds
.
10
12
37,502
Cause/Description
Failure of pump seal due to
freezing led to vapor release
of ethylene oxide; heat of
friction caused fire
Combination of extreme cold
temperatures and hot
operating temperatures led
to warping and rupture of
flange joints, resulting in
vapor release of ethylene
oxide and subsequent fire
Valve failure on process
vessel allowed atmosphere
above ethylene oxide liquid
to exceed lower explosion
limit, resulting in vapor
release, fire and explosion
Explosion occurred during
trial operation of new
process equipment while
reactants were being
removed. Explosion
damaged process vessels and
piping, releasing methyl
chloride gases
End Effects
Vapor release,
fire
Vapor release,
fire
Vapor release,
fire, explosion
Explosion, fire,
vapor release
Facility
Injuries
0
0
0
1
.
Deaths
0
0
0
0
Public
Injuries
0
0
0
0
•»
Deaths
0
0
0
0
Number
Evacuated
0
0
0
0
-------
-F-8-
Chemical (s)
Propylene and
Propane
Silane
•
Quantity
Released in
Pounds
1,000
.
*
Cause/Description
Operator error during
maintenance activity allowed
valve to leak propylene and
propane vapor and liquid,
which ignited leading to
subsequent fire and
explosions
Employees attempted to vent
a cylinder of compressed
silane gas contaminated with
nitrous oxide when it
exploded; led to subsequent
explosions and fire
destroying the entire facility
End Effects
Vapor release,
explosion, fire
Explosion, fire
Facility
Injuries
2
0
Deaths
0
, '
2
Public
Injuries
0
0
Deaths
0
0
Number
Evacuated
150
0
-------
-F-9-
Exhibit F-5 provides information about explosions in the OSHA data base. The OSHA data tends to
cover releases that have on-site rather than off-site consequences. In 1990, 25 employee deaths and 43 injuries we
reported for flammable chemical explosions. The causes of these accidents are mostly equipment failure including
general pipeline failure, valve failure, process unit failure, and leaking torches. In addition to equipment failure,
employee error or misjudgement of hazardous situations frequently contribute to an accident. Operators use
equipment that is inappropriate or do not follow standard operating procedures. Also, mistakes by maintenance
personnel often cause explosions. In many accidents, inadequate purging of vapors from tanks, wells, or vaults
have caused explosions with common ignitions sources including welding torches, gas heaters, pilot lights, lighters,
and cigarettes.
-------
-F-10-
Exhibit F-5
INFORMATION ABOUT EXPLOSION ACCIDENTS INVOLVING FLAMMABLE CHEMICALS
IN THE OSHA DATABASE
Chemical
acetylene
acetylene
aluminum dust
butane
butane
2-butanone
carbon
monoxide
flammable
liquid
containing
toluene and
naphtha
flammable
liquid
flammable
liquid
flammable
liquid
flammable
liquids
flammable
vapor cloud
Human Factor
insufficient or lack of
engineering controls
misjudgment of
hazardous situation
malfunction in
securing/warning
operation
misjudgment of
hazardous situation
butane lighter carried in
clothing
misjudgment of
hazardous situation
insufficient or lack of
engineering controls
insufficient/lack of
practice procedures
misjudgment of
hazardous situation
equipment inappropriate
for operation
misjudgment of
hazardous situation
misjudgment of
hazardous situation
unknown
Cause
generator lost power and acetylene gas
accumulated and ignited from a spark ,
from electric control room
employee filled bag with acetylene to
ignite as practical joke; static spark
ignited bag prematurely
welding sparks ignited dust and
aluminum fines creating fire ball
butane leak ignited
hot slag ignited butane lighter
employee purging spray paint gun on
top of drum left generator operating;
voltage from generator passed through
metal plumbing and ignited drum
gas heater ignited CO and other flue
gases
piece of slag or spark from welding
ignited vapors
gas cutting torch ignited tank previously
containing flammable liquid
gas space heater ignited spilled
flammable liquid
cigarette lighter ignited clothing
splashed with flammable liquid
electric arc welder ignited waste oil tank
with oil and flammable liquids
a released vapor cloud ignited and
exploded
Number
of
Injuries
1
2
7
1
0
2
0
0
1
0
2
1
3
Number
of
Deaths
0
0
0
1
1
1
1
1
0
1
1
1
2
-------
-F-ll-
Chemical
flammable
vapors
flammable
vapors
flammable
vapors
flammable
vapors
hydrogen
liquified
propane gas
liquified
propane gas
methane
methane
molten
aluminum
natural gas
natural gas
natural gas
propane
propane
Human Factor
used unapproved
equipment in flammable
atmosphere/misj udgment
of hazardous situation
equipment inappropriate
for operation
equipment inappropriate
for operation
misj udgment of
hazardous situation
misjudgment of
hazardous situation
malfunction in
securing/warning
operation
disconnected gas line
misjudgment of
hazardous situation
malfunction in
securing/warning
operation
malfunction in
securing/warning
operation
misjudgment of
hazardous situation
insufficient or lack of
practice procedures
misjudgment of
hazardous situation
malfunction in
securing/warning
operation
insufficient or lack of
practice procedures
Cause
electric skill saws ignited vapors in tank
static electricity ignited vented vapors
which caused a vessel explosion
gas-powered water heater ignited vapors
in spray paint booth
employee scraping underground storage
tank ignited vapors with oxygen-
acetylene torch
chemical reaction occurred within tank
producing hydrogen gas; ignited gas with
torch when opening tank
gas leak caught on fire
match ignited liquified propane gas from
disconnected line
cigarette lighter ignited natural gas
vapors from damaged utility line ..
open check valve allowed hydrocarbon
gas to escape and ignite
aluminum billet mold fell into coolant
and molten aluminum exploded
spark from engine ignited gases from oil
and natural gas well
leaking cutting torch ignited electrode
holder causing explosion
dozer blade ignited gas line
unrepaired gas leak in smokehouse pilot
light caused fire in smokehouse burner
match struck inside underground utility
vault ignited propane gas
Number
of
Injuries
1
0
1
0
1
0
1
3
8
1
1
0
0
1
1
Number
of
Deaths
1
1
0
1
2
1
0
0
1
0
4
1
1
0
1
-------
-F-12-
Chemical
propane
propane
propane
propane
Human Factor
misjudgment of
hazardous situation
misjudgment of
hazardous situation
safety devices
removed/inoperable
malfunction in
securing/warning
operation
Cause
cigarette ignited propane leaking from
tank
torch flamed out allowing unburned
propane to accumulate and explode
gas pocket formed and exploded when
pilot on grill was lit '
pilot light ignited gas escaping from
seemingly closed tank valve
Number
of
Injuries
1
0
1
2
Number
of
Deaths
0
1
0
0
-------
-F-13-
Exhibit F-6 cxivers flammable chemical accidents in the M & M data base. M & M lists dollar damage
estimates as well as most off-site damage. Off-site damage occurs less frequently than on-site damage and is more
difficult to characterize and measure. Most damage is measured in terms of windows broken. Flying shards of
glass would cause injuries and possible death to nearby public residents. In the most forceful explosions, windows
were broken 9.6 kilometers (six miles) away from the explosion center. From the M &• M data, it appears that
simple hydrocarbon chemicals are responsible for many of the explosions causing off-site damage. The reason may
be that these chemicals are produced in large quantities and when released, form detonable vapor clouds.
-------
-F-14-
EXHIBIT F-6
INFORMATION ABOUT EXPLOSION ACCIDENTS INVOLVING FLAMMABLE CHEMICALS IN THE M & M DATABASE
Chemical
Butane
Butane
Butane (liquid)
Cumene
Cumene
Cyclohexane
Ethane and
Propane
Elhylene
iHCiJeBl
Billings. MT (8/14/72)
Pampa.TX (11/14/87)
Feytin (Lyon), France
(1/4/66)
Deer Park. TX
(5/17/80)
Philadelphia, PA
(3/9V82)
Flixborough, UK
(6/1/74)
Baton Rouge
(12/24/89)
Pasadena, TX
(10/23/89)
Tfft tt Facility
Refinery
Petrochemical
Plant
Refinery
Petrochemical
Plant
Petrochemical
Plant
Refinery
Petrochemical
Plant
Ca»e/D«icrl»ll»
Valve on deisobutanizer blocked by
powder. When flange was opened,
powder blew out, releasing the liquid
butane.
Explosion in either an air pipeline
or a manifold at an acetic acid
production reactor led to release of
gas from ruptured pipe, resulting in
vapor cloud which exploded
Improper sampling procedures.
Seal on pump failed.
Upset temperature condition caused
cumene to be vented.
Failure of pipeline on bellows unit
Ruptured pipe due to cold weather
led to vapor cloud release, which
ignited resulting in vapor cloud
explosion
Release of gas (cause unknown) at
high pressure led Co formation of a
large vapor cloud which ignited,
resulting in a major initial explosion
and later multiple explosions
Damafe
{14 million
S24I.1M
S69 million
(26 million
S29 million
$44. 7M
SSOOto
S750M
C«BS«cjB«BCcs
OB-.H.
Drums destroyed, fracu'onation tower
fell across two major pipelines.
Destruction of most buildings and
heavy damage lo most process units
Five tanks destroyed, 40 lines cut
Fin-fan coolers collapsed; one process
column destroyed.
Three process tanks and one fuel oil
lank were destroyed.
Firestorm followed vapor cloud
explosion.
Fire damage to 2 large storage tanks,
12 small tanks; ruptured pipelines
facility shutdown for 3 days
Destruction of 2 major process units;
stoppage of production for 18 to 24
months
Off-lilt
Rupture of an underground
fire main
Damage to windows 6 miles
away; damage to electric,
steam, and fire water
supplies
•»
Other Factors A»«»l ExpUsUn
•Vapors ignited al furnace 100 feet from
point of release.
•Reactor maintained at 285* and 700 psi
•Ignition occurred 10 seconds after
release
•Ignition 300 feet from point of release
•Projectiles up to 100 tons travelled 3/4 .
of a mile
•Ignition occurred 35 minutes after
release
•60.000 pounds released at ISS'C
•Vapor cloud 500 ft x 400 ft
•Pipeline ruptured at 700 psi
•Ignited a few minutes after release
•3.6 million gallons of diesel fuel and
880,000 gallons of lube oil were ignited
•Elhylene released at 700 psi
•Ignition occurred one minuted after
release
•TNT equivalent of 10 tons
-------
-F-15-
Chemical
Ethylene
Oxide
Eihvtene
Oxide
Ethylene
Oxide.
Propylene
Oxide, Clycol
•nd Glycerin
Ethylene,
Propylene
Flammable
solvent
Gil
Gts (Fuel)
Hydrocarbon
liquid and
gases
ImcUnt
Doe, KY (4/17/62)
Port Lavaca. TX
March 12, 1991
Geismar, LA
(5/2V76)
Priola. llaly (S/19/8S)
Cincinnati. OH July
191990
Abqakj, Saudi Arabia
(4/15/78)
Abqak). Saudi Arabia
(6/4/77)
Beaumont, TX
(1V29/74)
Typ< .f FacWrjr
Petrocbemical
Complex (Elbylene
Oxide Plant)
Petrochemical
Plant
Petrochemical
Plant
Resin can-coatings
plant
Gas Processing
Plant
Gas Processing
Plant
Petrochemical .
Plant
C»<
-------
-F-16-
Ckimlcil
Hydrocarbons
Hydrocarbons
(liquid and
gaieous)
Hydrogen
Hydrogen •
Hydrogen and
Hydrocarbons
Isobuune
LPO
LPG
IncMeK
Pemis, Netherlands
(1/20/68)
Teas City. TX
(7/21/79)
, Grangemouth, United
Kingdom (3/22*87)
Richmond, CA
(4/10/89)
Martinez, CA (9/5/89)
Lake Charles. LA
(8/8/67)
Mexico City. Mexico
(11/19/84)
Rio de Janeiro, Brazil
(3,30/72)
Ty»« .f F.ellilj
Refinery
Refinery
Refinery
Refinery
Refinery
Refinery
Terminal
Refinery
Ca«t /Description
Runaway reaction in slop tank
resulted in boil over and vapor
release.
Reflux accumulator line failed.
Separator overpressurized, allowing
release through relief valve.
Failure of pipeline at a weld led to a
leak of gas, resulting in a high-
pressure fire
Failure of line downstream from a
separator at a hydrotreater unit led
to vapor cloud release, resulting in
fire or explosion
Repairs to corroded valve further
stressed tbe valve and increased the
lize of the leak; valve bonnet blew
oft.
Sphere ruptured while receiving
product; incoming flow could not be
ihut oh*.
Operator left LPG drain valve open.
Damage
$96 million
$37 million
$88 million
S93.6M
J52M
$63 million
$23 million
$13 million
C«*)>e{«eneei
Oi -file
Two wax cracking units, one naphtha
cracking unit, one sulfur plant, and
60 storage tanks were either destroyed
• or damaged.
Alkylation unit, FCC unit, and control
building sustained heavy structural
damage.
Separator exploded and disintegrated.
Loss of 25 percent of refinery's
throughput capacity; interruption of
gas oil hydrocracker operations for S
months
Destruction of nydrolreater unit
One alkylation unit destroyed Six
cells of main cooling tower wrecked '
Spheres ruptured, cone roof tanks
caught fire.
Four spheres and 44 bullets ruptured.
All 2 1 tanks were destroyed; refinery
facilities and buildings sustained blast
damage.
on;-«ite
• Damages reported up to
9V4 miles away
,
• Windows broken half
mile away
• Windows broken six
miles away
t*
Other Factors A»»l Ex>lxi»
•Between SO and 100 tons of
hydrocarbon slop boiled over
•Exploded with a long-range TNT
equivalent of 100 tons
•Cloud ignited 6-8 minutes after release
•4.000 to 5,000 gallons of liquids were
discharged
•Vapor cloud travelled 640 feet
•Ignition occurred 2 minutes after initial
release
•Projectile weighing 3 tons thrown 3,300
feet
•Affected line at 2,800 psi
•17,500 gallons of isohutane released
•Vapor cloud size estimated at 300 feet
x 800 feet x 20 feet
•Ignition 10-15 minutes after spill
•Product received at 341 psig
•Initial ignition occurred 10 minutes
after vapor cloud formed
•Five minutes after initial ignition, series
of BLEVEa erupted
•Fireball estimated to be 1.200 (eel in
diameter •
•Projectiles weighing 20 Ions were
thrown 3,900 feet
•LPG initially released at 1$6 psi; relief
valve release occurred at 239 psi
-------
-F-17-
Chemical
Methyl-
lertiary-butyl-
ether,
propylene
oxide, or
styrene
Nitromethane
Propane
Propane
Propane
Propane
Propane
Propane
IncIJtil
Channelview, TX July
5. 1990
Surlinglon, LA
May 1. 1991
Borger. TX (1/20/80)
Denver, CO (10/3/78)
Moni BeMeu, TX
(11/5/85)
Port Arthur, TX
(6/8/88)
Port Hudson, MO
(119/70)
Rai Tanura, Saudi
Arabia (8/15/87)
Type .r Facility
Petrochemical
Plant
NitroparaRin Plant
Refinery
Refinery
Terminal
Refinery
Ga> Processing
Plant
C»«/Dt>criplU«
Explosion in giant holding tank
occurred during maintenance.
Source of ignition is unknown.
Fire near a compressor detonated
nitro methane.
Piping/vessel overpressured after
plug froze on contact with propane.
Pipe on stabilizer reboiler failed.
Human error resulted in breaking
of a high-pressure gas pipeline,
leading to vapor doud release, with
• subsequent explosion and Tire
Major failure of pipeline led to
' vapor doud release; ignition and
resulting explosion led to additional '
pipe failures and Ores
Pipeline ruptured
Release attributed to a flange in a
relief valve line.
Damat*
$90 million
$110
million
$49 million
$37 million
J44.8M
S17.4M
KJ miUion
C**s«4«cneec
O»->llc
Deaths included five ARCO
employees, one independent truck
driver, and 1 1 non-union employees.
The explosion leveled an area the size
of a city block.
Severely damaged the Angus facility.
Debris included masses of twisted
metal tanks and pipes.
Alkylation unit and boiler plant
destroyed. Refinery shut down.
Catalytic polymerization unit
destroyed, other refining units heavily
damaged.
Total loss of electronic equipment and
computers in fire which radiated heat
and melted glass windows of facility;
large loss of gas inventory
Heavy damage to storage tanks and
pipelines
Complete destruction of warehouse
OITslh
•
• Destroyed much of the
town's main business
district.
• 20 families were left
homeless.
• Homes, hospital, and
commerdal buildings were
damaged.
• Damage induded broken
windows, collapsed ceilings,
fallen siding from houses,
bumed-out vehides, and
Utter of twisted metal
•Broken windows up to 5
miles away
»
Olhtr FacUn Al»«l EipU>l»
•The fire sent a huge doud of black
smoke into the air that was visible 6
miles from the Mast.
•The douds of smoke moved away from
residential areas.
•500-600 residents were evacuated from
the immediate area
•1 to 2 minutes between "pop" on
recycle compressor and detonation
•Vapor doud ignited approximately 300
feet from point of release
•Line under 800 psi
•Vapor doud size estimated at 44 acres
•Ignition occurred 4-5 minutes after
release
•132,000 pounds of propane released at
942psig
•Vapor doud covered 10 acres
•Ignition occurred after 24 minutes
•1,900 barrels of propane were released
-------
-F-18-
Cktnlc.l
Propane
Propane
Propane
Propane and
Propylene
Propane,
Butane
Propane
(C3 Hydrocarb
ont)
Propylene
lncMenl
RomeoviHe, IL
(7/23/84)
Torrance, CA
(11/24/87)
Wood River (1/23/8$)
Morrii (S/7/W)
Linden, NJ (3/20/79)
Norco (5/5/88)
East St. Louis. IL
(1/22/72)
Ty»« >r Fa.llily
Refinery
Refinery
Refinery
Petrochemical
Plant
Refinery
Refinery
C»ie/D»cri»ll»
Cracked circumferential weld leaked
propane.
Failure of treater in a refinery
process unit exploded into and
tevered pipeline* and flarelinet,
mulling in multiple firet
Ruptured pipe due to cold weather
led to vapor cloud release, which
ignited resulting in vapor cloud
explosion
Power outage and operator error in
venting propytene caused vapor
cloud
Dead-end section of piping in FCC
unit failed.
Internal corrosion in pipeline caused
failure in pipeline elbow, which kd
to a vapor cloud release resulting in
a major explosion with damage both
on and off-site
Railroad can collided and propylene
released into vapor cloud; a second
explosion also occurred.
Dama£c
$144
million
S16.8M
S25.2M
S41.6M
$27 million
S327M
Cvaseqvtiicts
Oi-iltt
MEA absorber column exploded,
toppled power transmission tower.
Pipelines sheared, water treatment
tank ruptured.
Extensive damage to the HF aikylation
unk stripper, alumina treater, and
depropanizer column
Unspecified
Extensive damage to 40 acres of
production facility, Including the
ethylene production area
Unused control room destroyed;
debris severed lines, releasing
hydrocarbons.
Severe and extensive damage to
majority of refinery; fluid catalytic
cracking unit was demolished and new
one will be built
OR-ille
•Broken windows up to f
miles away
,
5,200 property claims
received for damages up to
6 miles away
•School, located 1 kitomrter
from explosion, suffered
internal damage.
Olfctr Facttra A»..l EI»|«S!M
•Propane released ai 200 psi, 100* F
•Projectile weighing 20 tons thrown
3,500 feel
•500 barrels of propane-butane mix were
released as a result of initial explosion
•BLEVE resulted in one projectile
travelling 500 feet and another travelling
<00 feet
•Vapor doud 1.5 acres, 5-6 feet deep
•20,000 pounds of propane were released
•Ignition occurred 30 seconds after
release
•Depropanizer column at 270 psi and
130- F
•Overpressures up to 10 psi withing 100
meters
•53.5 tons released
•TNT equivalent (long-range) of 1-15
tons
•Vapor doud covered 5 acres
•Ignition 5 minutes after release
-------
-F-19-
Chemical
Propylene
Propylene,
hexane
Propytene
(other
hydrocarbons)
Styrene, vinyl
chloride
monomer
Vinyl,
ethylacerylene*
IncMeil
New Castle, DE
(10/21/80)
Goi, Japan (10/8/73)
Beet, Netherlands
(11/7/75)
LaSalle, Quebec,
Canada (IO/13/««)
Texas City, TX
(10Y23/W)
Tv§* .f F.clllly
Petrochemical
Plant
Petrochemical
Plant
Petrochemical
Plant
Petrochemical
Plant
Petrochemical
Plant
Ca«se/D«ccri>tl«>
Operators removed valve.
Operator opened wrong valve.
releasing chemicals.
Leak caused by cold brittle fracture
of a feed drum connection to the
safety valve.
Reaction in mass polymerization out
of control; rupture disc relieved.
Heat-triggered decomposition.
Damage
J64 million
$19 million
S47 million
111 million
$27 million
CcMC^VCBCCB
On-flle
Two process lines, the control building
and the finishing area were severely
damaged.
Storage tanks and buildings destroyed.
A polystyrene building, an electric
substation, and a warehouse destroyed.
Rail car was knocked over.
Five towers destroyed or seriously
damaged.
Olt-tiU
*
•2,310 cases of off-site
damages
Other Factors A».«l ExpUtUa
•12,000 to 16,000 pounds of monomer
released at ISO pii
•Vapor cloud, 2SO feet x 450 feet
•Ignited after l.S-2 minutes
•Vapor cloud ignited by a relay on an
extruder ISO feel from point of release
•Three to five tons of hydrocarbon mist
were released
•Vapor cloud 100 meters in diameter, 1-
2 meters high
•Igniliom 2 minutes after release
•TNT equivalent of 59 tons
•1,500 Ibs. of styrene were released and
ignited
•22,000 gallons of vinly monomer and
60,000-80.000 Ibs. of additional styrene
were ignited
•540 gallons of liquid hold-up ignited
•Projectile of 600 Ibs. travelled 3,000 feet
-------
-G-l-
APPENDIX G
CONTACTS WITH EXPERTS
Several experts were contacted for recommendations on models for flammables and for information
on other issues related to flammables, such as vapor cloud explosions, accident scenarios, and potential
consequences. Information acquired through discussions is presented for each expert individually.
Bob Benedetti
Flammable Liquids Engineer
NFPA
Mr. Benedetti suggested that models are an inappropriate method for examining the consequences
of a spill or pool fire on a community. He said that pool fires rarely leave the fence line of the facility.
Mr. Benedetti said that NFPA maintains a database on fires but a search for explosions is too
general for the database. He also suggested that NFPA 30, the Combustible Liquid Code, might provide a
good source of information for examining flammables. He said that these codes are basis for law in two-
thirds of the states.
Dr. John Boccio
Department of Nuclear Energy
Brookhaven National Laboratory
Dr. Boccio felt that one area that perhaps could be covered in somewhat more depth is, regarding
flammable gases and liquids, the effects of propagation of the explosion into an area with obstacles, i.e.,
consequences of explosions on material structures, such as buildings. He said that we could probably find
references in the literature to explosions in mine shafts.
Regarding the need to further validate the models of explosions included in our report, such as
WHAZAN, etc., Dr. Boccio's view is that these models have been developed by established and credible
organizations, and are widely used in industry, hence he wonders if the effort expended in further
validating them would be a worthwhile one.
He also mentioned that DOE has shown an interest in developing their understanding of the
hazards posed by explosive, flammable, and reactive chemicals, and that perhaps, at some point in the
future, both the interests of EPA and DOE might be served by some sort of collaborative effort on this
topic. He referred us to work being done in DOE's Office of Environmental Health.
-------
-G-2-
David V. Eberhardt
Senior Engineer and Hazard Analyst
Rohm & Haas Co.
Mr. Eberhardt was very well acquainted with the operation of the PHAST model for performing
consequence analysis for flammable chemicals. He suggested that PHAST is the best model for screening
hazards, and examining the impact of preventable releases such as line ruptures, accidental releases,
BLEVE's, leaks and vents.
The PHAST model is a user friendly, PC based, modeling program designed jointly by Rohm &
Haas Co. and Technica International, LTD of Columbus, Ohio at( 614)848-4000 or (7ll)447-9400 in Los
Angeles. Mr. Eberhardt said that Rohm & Haas Co. uses the PHAST model to screen accident scenarios.
If by using the PHAST model the accident is shown to go beyond the fence line, the accident scenario is
examined further using a sophisticated risk assessment model operated on a mainframe.
The PHAST model is easily operated. It is a menu driven program which also includes the ability
to input specific parameters to tailor the model to the user's needs. Mr. Eberhardt said that Rohm &
Haas Co. has used the PHAST model for vapor cloud dispersion, vapor cloud fires, and vapor cloud
explosions (VCEs). It has been used to study the velocity of atomized material and travel distances for
droplets. One useful feature that Mr. Eberhardt suggested was the coupling of accident scenarios (e.g.,
BLEVE's) which can be done with the PHAST model.
Mr. Eberhardt suggested that a Technica International user's group would be an excellent source of
information regarding the operation of the PHAST model. Since Technica International operates a larger
consequence model, updates are made to the PHAST program first, and general comments regarding the
changes are solicited. This ensures that the PHAST model is extremely-up-to-date. Mr. Eberhardt said
that changes are made to the PHAST model about twice a year.
Mike Johnson
PHAST Program Manager
Technica International
Mr. Johnson answered questions related to the PHAST model, including technical questions related
to model operation, source of the equations used in the models, and overpressure levels used PHAST. He
also discussed the differences between PHAST and WHAZAN.
Dr. Marvin D. McKinley
Professor of Chemical Engineering
University of Alabama
As far as a state-of-the-art model for performing consequence analysis for flammable chemicals, Dr.
McKinley has heard positive feedback about the PHAST model developed by Technica International and
Rohm & Haas.
-------
-G-3-
Dr. McKinley is an expert on explosions of LPG and natural gas. He believes that EPA analysis
should focus on the liquified gases as a source of VCE. Also, EPA assumptions in the consequence
analysis need to reflect reality. NFPA and several states have regulations and codes that begin to address
the dangers of VCE. He feels that industry follows these codes and regulations and that the specifications
found in these codes (e.g., separation distances) should form the basis of EPA's assumptions in the
consequence analysis. For example, if an NFPA code specifies that propane tanks should be 50 feet away
from the property line, EPA should measure all overpressures at 50 feet. Dr. McKinley also feels that
EPA needs to consider the conditions under which the flammables are used and stored in order to
estimate the possible release quantities of flammable gases.
Concerning overpressures, Dr. McKinley agrees with Dennis Wade from Monsanto tnat EPA must
first consider overpressures that can damage buildings - and thereby kill and injure people - before
considering overpressures that can directly harm people. Concerning thermal exposure assumptions, Dr.
McKinley believes that there needs to be a better combination of time and exposure for BLEVEs. Also,
EPA should look into diking requirements when examining pool fire scenarios.
The rocketing of containers (e.g., projectiles) was not a great concern for Dr. McKinley. Even with
BLEVEs, he felt that firefighters have enough time to cool the vessel thereby preventing the possibility of
projectiles from an explosion. In fact, he doesn't see much chance of BLEVEs today.
Concerning accident scenarios, Dr. McKinley has found that most explosions of flammable
chemicals result from a worker failing to follow procedures or ignoring existing information about hazards.
For example, many explosions occur when someone fails to inert a tank previously filled with an
flammable chemical and another person attempts to weld the tank. For additional scenarios, Dr. McKinley
suggests we consult NFPA's Fire Journal Magazine for descriptions of incidents involving flammable
chemicals. Also, he suggests we contact Dr. Grelecki for information on test data.
Dr. Dennis Wade
Manager of Safety Technology (Retired)
Monsanto Co.
Dr. Wade has specific concerns that he believes were not adequately addressed in the Science
Advisory Board Meeting. Dr. Wade feels that the greatest danger from flammable chemicals is the threat
of explosions and not fires. Nobody has died in the community from thermal radiation resulting from
ignition of flammables. However, the overpressure from a vapor cloud explosion may cause death or
injury to the community. Also, he believes that EPA should focus on overpressures that cause a building
to collapse rather than on overpressures that directly result in a death or injury. Historically, secondary
effects such a building collapse have caused more deaths and injuries. Dr. Wade used the explosion in
Pampa, Texas to support his opinion.
Dr. Wade recognized that there is a problem in trying to relate quantity of a release of flammable
chemicals and the overpressure created by the explosion. He has some reservations about the WHAZAN
and ARCHIE models because of the assumptions in those models. He also disagrees with EPA's
consideration of 50% lower flammability limit. To gather information on state-of-the-art models for
consequences of flammable explosions, Dr. Wade suggests three organizations:
AIChE Center for Chemical Process Safety - a committee of experts is working on models and
guidelines in the area of vapor cloud explosions
-------
-G-4-
Bureau of Industrial Safety (TNO) of the Netherlands - reissued the Yellow Book
Christian Michelson Institute -
Dr. Wade suggests that the models need to be reality tested with credible input and validated with
test data. Examining accident scenarios can be helpful. Also, he suggests that EPA contact John
Davenport, a world expert on data from vapor cloud explosions. Mr. Davenport is with Industrial Risk
Insurers (Hartford).
The flammable chemicals considered in the consequence model should be those that are volatile at
atmospheric pressure (e.g., propane, propylene). Other chemicals that must be heated to volatilize are less
important to consider.
The trigger quantity must be large enough to create a vapor cloud. Monsanto used 15,000 pounds
as a danger quantity, however, based on new data, the quantity has been reduced to 7,000 pounds.
Dr. Wade believes that one important scenario has not been adequately addressed - flammables in a
closed building. This scenario combines the hazards of concentrated flammable vapors in a confined space,
the possible large number of persons working in a building, and the issue of projectiles. In addition, EPA
should look into the dangers posed by flammables at the retail and distributing level, rather than just the
chemical manufacturing level.
Dr. Chester Grelecki
Hazards Research Corporation
Dr. Grelecki presented an overview of fire and explosion hazards at a meeting at EPA. He also
reviewed documents related to fire and explosion hazards, as well as answering questions and providing
information to EPA in a conference telephone call discussing fire and explosion hazards.
-------
50272 -101
REPORT DOCUMENTATION
PAGE
l._ REPORT NO.
EPA 744-R-94-002
3. Recipient's Accession No.
4. Title and Subtitle
Flammable Gases and Liquids and Their Hazards
5. Report Date
Feb 1994
Authorts)
8. Performing Organization Rept. No.
9. Performing Organization Name and Address
ICF Inc.
9300 Lee Hwy
Fairfax, VA 22031
10. Project/Task/Work Unit No.
Work Assignment #26
11. Contract(C) or Grant(G) No.
to 68-C2-0107
(G)
12. Sponsoring Organization Name and Address
U. S. Environmental Protection Agency
Office of Pollution Prevention and Toxics (7406)
401 M. St. S.W.
Washington, D.C. 20460
13. Type of Report & Period Covered
Final
14.
IS. Supplementary Notes
16. Abstract (Limit: 200 words)
This report assesses the potential consequences of accidents involving flammable
chemicals. The analysis includes the identification and evaluation of existing
listing and classification systems, along with any applicable criteria; review of
existing regulations and codes dealing with flammable materials; analysis of
histories of accidents involving flammable substances; and modeling potential
consequences of fires and explosions of flammable substances. The results of
this report may be used to support regulatory and non-regulatory objectives of EPA
under the Superfund Amendments and Reauthorization Act (SARA) Title III 1986; and
the Clean Air Act Amendments (CAAA) 1990.
17. Document Analysis a. Descriptors
Flammable
Hazard
b. Identifiers/Open-Ended Terms
iSATI Field/Group
Statement
Release unlimited
L
19. Security Class (This Report)
20. Security Class (This Page)
21. No. of Pages
172
22. Price
(SeeANSI-Z39.18)
See Instructions on Reverse
OPTIONAL FORM 272 (4-77)
(Formerly NTIS-3!)
Department of Commerce
------- |