Chapter 1 - Chemical Process QRA
Chapter 1 - Chemical Process QRA
Chapter 1 - Chemical Process QRA
• Chemical reactions may be involved
• Processes are generally not standardized
• Many different chemicals are used
• Material properties may be subject to greater uncertainty
• Parameters, such as plant type, plant age, location of surrounding population,
degree of automation and equipment type, vary widely
• Multiple impacts, such as fire, explosion, toxicity, and environmental contamina
tion, are common.
Many hazards may be identified and controlled or eliminated through use of quali
tative hazard analysis as defined in Guidelines for Hazard Evaluation Procedures., Second
Edition (CCPS, 1992). Qualitative studies typically identify potentially hazardous
events and their causes. In some cases, where the risks are clearly excessive and the exist
ing safeguards are inadequate, corrective actions can be adequately identified with
qualitative methods. CPQRA is used to help evaluate potential risks when qualitative
methods cannot provide adequate understanding of the risks and more information is
needed for risk management. It can also be used to evaluate alternative risk reduction
strategies.
The basis of CPQRA is to identify incident scenarios and evaluate the risk by defin
ing the probability of failure, the probability of various consequences and the potential
impact of those consequences. The risk is defined in CPQRA as a function of probabil
ity or frequency and consequence of a particular accident scenario:
Risk = F(s, c,f)
s = hypothetical scenario
c = estimated consequence(s)
/ = estimated frequency
This "function" can be extremely complex and there can be many numerically dif
ferent risk measures (using different risk functions) calculated from a given set of s, c,f.
The major steps in CPQRA, as illustrated in Figure 1.1 (page 4), are as follows:
Risk Analysis:
1. Define the potential event sequences and potential incidents. This may be based
on qualitative hazard analysis for simple or screening level analysis. Complete
or complex analysis is normally based on a full range of possible incidents for all
sources.
2. Evaluate the incident outcomes (consequences). Some typical tools include
vapor dispersion modeling and fire and explosion effect modeling.
3. Estimate the potential incident frequencies. Fault trees or generic databases
may be used for the initial event sequences. Event trees may be used to account
for mitigation and postrelease events.
4. Estimate the incident impacts on people, environment and property.
5. Estimate the risk. This is done by combining the potential consequence for each
event with the event frequency, and summing over all events.
Risk Assessment:
6. Evaluate the risk. Identify the major sources of risk and determine if there are
costeffective process or plant modifications which can be implemented to
reduce risk. Often this can be done without extensive analysis. Small and inex
pensive system changes sometimes have a major impact on risk. The evaluation
may be done against legally required risk criteria, internal corporate guidelines,
comparison with other processes or more subjective criteria.
7. Identify and prioritize potential risk reduction measures if the risk is considered
to be excessive.
Bisk Management:
Chemical process quantitative risk analysis is part of a larger management system.
Risk management methods are described in the CCPS Guidelines for Implementing Pro
cess Safety Management Systems (AIChE/CCPS, 1994), Guidelines for Technical Manage
ment of Chemical Process Safety (AIChE/CCPS, 1989), andPtow* Guidelines for Technical
Management of Chemical Process Safety (AIChE/CCPS, 1995).
The seven steps in Figure 1.1 are typical of CPQRA. However, it is important to
remember that other risks, such as financial loss, chronic health risks and bad publicity,
may also be significant. These potential risks can also be estimated qualitatively or
quantitatively and are an important part of the management process.
This chapter provides general outlines for the major areas in CPQRA as listed
below. The subsequent chapters provide more detailed descriptions and examples.
1. Definitions of CPQRA terminology (Section 1.1)
2. Elements that form the overall framework (Section 1.2)
3. Scope of CPQRA (Section 1.3)
4. Management of incident lists (Section 1.4)
5. Application of CPQRA (Section 1.5)
6. Limitations of CPQRA (Section 1.6)
7. Current practices (Section 1.7)
8. Utilization of CPQRA results (Section 1.8)
9. Project management (Section 1.9)
10. Maintenance of study results (Section 1.10)
CPQRA provides a tool for the engineer or manager to quantify risk and analyze
potential risk reduction strategies. The value of quantification was well described by
Lord Kelvin. Joschek (1983) provided a similar definition:
a quantitative approach to safety . . . is not foreign to the chemical industry. For every
process, the kinetics of the chemical reaction, the heat and mass transfers, the corrosion
rates, the fluid dynamics, the structural strength of vessels, pipes and other equipment
as well as other similar items are determined quantitatively by experiment or calcula
tion, drawing on a vast body of experience.
CPQBJV enables the engineer to evaluate risk. Individual contributions to the
overall risk from a process can be identified and prioritized. A range of risk reduction
measures can be applied to the major hazard contributors and assessed using
costbenefit methods.
Comparison of risk reduction strategies is a relative application of CPQRA. Pikaar
(1995) has related relative or comparative CPQRA to climbing a mountain. At each
stage of increasing safety (decreasing risk), the associated changes may be evaluated to
see if they are worthwhile and costeffective. Some organizations also use CPQRA in
an absolute sense to confirm that specific risk targets are achieved. Further risk reduc
tion, beyond such targets, may still be appropriate where it can be accomplished in a
costeffective manner. Hendershot (1996) has discussed the role of absolute risk guide
lines as a risk management tool.
CPQRA Steps
Define the potential
accident scenarios
Evaluate the event Estimate the potential
consequences accident frequencies
Estimate the
event impacts
Estimate the
risk
Evaluate the
risks
Identity and prioritize
potential risk reduction
measures
Application of the full array of CPQRA techniques (referred to as component
techniques in Section 1.2) allows a quantitative review of a facility's risks, ranging from
frequent, lowconsequence incidents to rare, major events, using a uniform and consis
tent methodology. Having identified process risks, CPQRA techniques can help focus
risk control studies. The largest risk contributors can be identified, and recommenda
tions and decisions can be made for remedial measures on a consistent and objective
basis.
Utilization of the CPQRA results is much more controversial than the methodol
ogy (see Section 1.8). Watson (1994) has suggested that CPQRA should be consid
ered as an argument, rather than a declaration of truth. In his view, it is not practical or
necessary to provide absolute scientific rigor in the models or the analysis. Rather, the
focus should be on the overall balance of the QBA and whether it reflects a useful mea
sure of the risk. However, Yellman and Murray (1995) contend that the analysis
"should be, insofar as possible, true—or at least a search for truth." It is important for
the analyst to understand clearly how the results will be used in order to choose appro
priately rigorous models and techniques for the study.
100lb/min etc.
Release of Jet Fire
HCN from
a Tank Vent
Tank Full
BLEVE of Tank 50% Full
HCN Tank
etc.
After 15 min. Release
Unconfined Vapor After 30 min. Release
Cloud Explosion After 60 min. Release
etc.
FIGURE 1.2. The relationship between incident, incident outcome, and incident outcome
cases for a hydrogen cyanide (HCN) release.
The event tree in Figure 1.2 has been provided to illustrate the relationship
between an incident, incident outcomes, and incident outcome cases. Each of these
terms will be developed further in this chapter.
It is convenient (for ease of understanding and administration) to divide the complete
CPQRA procedure into component techniques (Section 1.2.1). Many CPQRAs do
not require the use of all the techniques. Through the use of prioritized procedures
(Section 1.2.2), the CPQRA can be shortened by simplifying or even skipping certain
techniques that appear in the complete CPQRA procedure.
1.2.1. Complete CPQRA Procedure
A framework for the complete CPQRA methodology for a process system is given in
Figure 1.3. This diagram shows
1. CPQRA Definition
2. System Description
3. Hazard Identification
4. Incident Enumeration
5. Selection
6. CPQRA Model Construction
7. Consequence Estimation
8. Likelihood Estimation
9. Risk Estimation
10. Utilization of Risk Estimates
A brief account of the role of each of the techniques is given below, and more
detailed accounts are given in the sections indicated.
SYSTEM DESCRIPTION EXTERNAL DATA SOURCES (§5)
HAZARD IDENTFICATION
(HEP Guidelines)
INCIDENT ENUMERATION
(§1.4.1)
LEGEND
Methodology Execution
INCIDENT Sequence
SELECTION (§1.4.2) Information Flow
Sequence
CPQRA MODEL
CONSTRUCTION (§1.22)
Reliability data (§5.5)
ECONOMIC ASSESSMENT Not
Acceptable
SYSTEM COST EVALUATION RISK ESTIMATION
CALCULATION QUALITY
USER REACTION ! Risk calculation Risk uncertainty,
SYSTEM MODIFICATION
(§4.4) sensitivity and
Frequency MODIFICATIONS MENU
importance (§4.5)
Reduction Reduction (D System
@ CPQRA
Design Control © Requirements
Inventory Operation Q Siting Acceptable
Layout Management
© Business strategy UTILIZATION OF RISK ESTIMATE
Isolation procedures NEW/MODIFIED
Not Risk Assessment
SYSTEM DESIGN
(Absolute or Relative)
acceptable (§18)
COMPLETE
Risk targets
REVISE BUSINESS STRATEGY
• ABANDON PROJECT
• SHUT DOWN OPERATIONS
(see above) (see above)
STEP 3 EXPERIENCE, CODES
IDENTIFY HAZARDS CHECKLISTS, HAZOPS, ETC.
STEP 4 LIST OF
ENUMERATE INCIDENTS ENUMERATED INCIDENTS
LIST OF SELECTED
STEPS INCIDENTS, INCIDENT
SELECT INCIDENTS OUTCOMES. INCIDENT
OUTCOME CASES
DESIGN ACCEPTABLE
CONSEQUENCE AND STEP 6 (CONSEQUENCES ACCEPTABLY
EFFECT MODELS, ESTIMATE CONSEQUENCES LOW AT ANY FREQUENCY
DECISION CRITERIA OF OCCURRENCE)
CONSEQUENCES ARE TOO HIGH
YES STEP?
MODIFY SYSTEM TO
REDUCE CONSEQUENCES
NO
HISTORICALANALYSIS DESIGN ACCEPTABLE
FAULT TREE ANALYSIS STEPS (FREQUENCIES ACCEPTABLY
EVENTTREE ANALYSIS ESTIMATE FREQUENCIES LOW FOR ANY CONSEQUENCES)
DECISION CRITERIA
FREQUENCIES ARE TOO HIGH
YES STEP 9
MODIFY SYSTEM TO
REDUCE FREQUENCIES
STEP 10 DESIGN ACCEPTABLE
COMBINE FREQUENCIES (COMBINATION OF
DECISION CRITERIA AND CONSEQUENCES TO CONSEQUENCES AND
ESTIMATE RISK FREQUENCIES
ACCEPTABLY LOW)
RISKSARETOOHIGH
YES STEP 11
MODIFY SYSTEM TO LEGEND
REDUCE RISK METHODOLOGY EXECUTION
NO SEQUENCE
INFORMATION FLOW
SEQUENCt
DESIGN UNACCEPTABLE
(COMBINATION OF CONSEQUENCES AND
FREQUENCIES UNACCEPTABLY HIGH)
FIGURE 1.4. One version of a prioritized CPQRA procedure.
• Step 10 Combine Frequency and Consequences to Estimate Risk. If the risk
estimate is at or below target or if the proposed strategy offers acceptable risk
reduction, the CPQRA is complete and the design is acceptable.
• Step 11 Modify System to Reduce Risk. This is identical in concept to Steps 7
and 9. If no modifications are found to reduce risk to an acceptable level, then
fundamental changes to process design, user requirements, site selection, or
business strategy are necessary.
1.3.1 The Study Cube
CPQRAs can range from simple, "broad brush" screening studies to detailed risk anal
yses studying large numbers of incidents, using highly sophisticated frequency and
consequence models. Between these extremes a continuum of CPQRAs exists with no
rigidly defined boundaries or established categories. To better understand how the
scope ranges for CPQRAs it is useful to show them in the form of a cube, in which the
axes represent the three major factors that define the scope of a CPQRA: risk estima
tion technique, complexity of analysis, and number of incidents selected for study. This
arrangement also allows us to consider "planes" through the cube, in which the value of
one of the factors is held constant.
1.3.1.1. THE STUDY CUBE AXES
For this discussion, each axis of the Study Cube has been arbitrarily divided into three
levels of complexity. This results in a total of 27 different categories of CPQRA,
depending on what combinations of complexity of treatment are selected for the three
factors. Each cell in the cube represents a potential CPQBA characterization. How
ever, some cells represent combinations of characteristics that are more likely to be
useful in the course of a project or in the analysis of an existing facility.
Intermediate
Simple
ORIGIN
Cube's
Bounding Main
INCREASING NUMBER OF
SELECTED INCIDENTS
Group Diagonal
Representative
Set
Expansive
List
FIGURE 1.5. The study cube. Each cell in the cube represents a particular CPQRA study with a
defined depth of treatment and risk emphasis. For orientation purposes, the shaded cells
along the main diagonal of the cube are described in Table 1.5.
• the complexity of the models to be used in a study
• the number of incident outcome cases to be studied
Model complexity can vary from simple algebraic equations to extremely complex
functions such as those used to estimate the atmospheric dispersion of dense gases. The
number of incident outcome cases to be studied is the product of the number of inci
dent outcomes selected and the number of cases to be studied per outcome. The
number of cases to be studied may range from one—assuming uniform wind direction
and a single wind speed—to many, using various combinations of wind speed, direc
tion, and atmospheric stability for each incident outcome.
Figure 1.6 illustrates how model complexity and the number of incident outcome
cases are combined to produce the simple, intermediate, and complex zones in the
study cube.
Number of Incidents. The three groups of incidents used in Figure 1.5—bounding
group, representative set, and expansive list—can be explained using the three classes of
incidents in Table 1.3.
The bounding group contains a small number of incidents. Members of this group
include those catastrophic incidents sometimes referred to as the worst case. The intent
of selecting incidents for this group is to allow determination of an upper bound on the
estimate of consequences. This approach focuses attention on extremely rare incidents,
rather than the broad spectrum of incidents that often comprises the major portion of
the risk. The representative set can contain one or more incidents from each of the
three incident classes in Table 1.3 when evaluating risks to employees. When evaluat
ing risk to the public, the representative set of incidents would probably only include
selections from the catastrophic class of events because small incidents do not normally
have significant impact at larger distances. The purpose of selecting representative inci
dents is to reduce study effort without losing resolution or adding substantial bias to
the risk estimate. The expansive list contains all incidents in all three classes selected
through the incident enumeration techniques discussed in Section 1.4.1.
SIMPLE/
ELEMENTARY INTERMEDIATE INTERMEDIATE
OF MODELS
SOPHISTICATED INTERMEDIATE/
INTERMEDIATE COMPLEX
FIGURE 1.6. Development of complexity of study axis values for the Study Cube. The main
diagonal values (shaded cells) correspond with the "complexity of study values" used in
Figure 1,5.
1.3.1.2. PLANES THROUGH THE STUDY CUBE
The study cube provides a conceptual framework for discussing factors that influence
the depth of a CPQRA. It is arbitrarily divided into 27 cells, each defined by three fac
tors, and qualitative scales are given for each factor or cube axis.
In addition to considering cells in the study cube, it is convenient to refer to planes
through the cube, especially through the risk estimation technique axis. A separate
plane exists for consequence, frequency, and risk estimation. Anywhere within one of
these planes, the risk estimation technique is fixed. Referring to consequence plane
studies, there are nine combinations of the complexity of study and number of selected
incidents. The use of the plane concept when describing CPQRAs is intended to rein
force the notion that several degrees of freedom exist when defining the scope of a
CPQRA study, and it is not enough to cite only the risk estimation technique to be
used when discussing a specific level of CPQRA.
1.4.1. Enumeration
The objective of enumeration is to identify and tabulate all members of the incident
classes in Table 1.3, regardless of importance or of initiating event. In practice, this can
never be achieved. However, it must be remembered that omitting important incidents
from the analysis will bias the results toward underestimating overall risk.
The starting point of any analysis is to identify all the incidents that need to be
addressed. These incidents can be classified under either of two categories, loss of con
tainment of material or loss of containment of energy. Unfortunately, there is an infi
nite number of ways (incidents) by which loss of containment can occur in either
category. For example, leaks of process materials can be of any size, from a pinhole up
to a severed pipe line or ruptured vessel. An explosion can occur in either a small con
tainer or a large container and, in each case, can range from a small ccpufP to a cata
strophic detonation.
TABLE 1.5. Definitions of Cells Along the Main Diagonal of the Study Cube (Figure 1.5)
Simple/Consequence CPQRA
Estimation Technique—Consequence
Complexity of Study
Number of Incident Outcome Cases—Small
Complexity of Model—Elementary
Number of Incidents—Bounding Group
This is a CPQRA that is useful for screening or risk bounding purposes. It requires the least amount
of process definition and makes extensive use of simplified techniques. In terms of Figure 1.4, it
consists of consequence calculations only (Steps I through 7). A Simple/Consequence CPQRA is
suitable for screening at any stage of the project: in the case of an existing plant, screening might
highlight the need to consider further study; at the design stage, it might aid in optimizing siting and
layout.
Intermediate/Frequency CPQRA
Estimation Technique—frequency
Complexity of Study
Number of Incident Outcome Cases—Medium
Complexity of Model—Advanced
Number of Incidents—Representative Set
This is a more detailed CPQRA that corresponds to Steps I through 9 in Figure 1.4. It cannot be
applied until the design is substantially developed, unless historical frequency techniques are applied. It
may be applied at any time after process flow sheet definition. Complete descriptions of the process
and equipment are not usually necessary. A Representative Set of incidents is chosen. In principle, the
results of an Intermediate/Frequency CPQKA should approximate a detailed study, but have less
resolution.
Complex/Risk CPQRA
Estimation Technique—Risk
Complexity of Study
Number of Incident Outcome Cases—Large
Complexity of Model—Sophisticated
Number of Incidents—Expansive List
This is the most detailed CPQRA. It employs the full methodology described in Figure 1.4. It may be
applied to operating plants or to capital projects, but only after detailed design has been completed,
when sufficient information is available. Where appropriate, it would employ the most sophisticated
analytical techniques reviewed in Chapters 2 and 3. However, it would be unlikely to apply the most
sophisticated techniques to all aspects of the study—only to those items that contribute most to the
result. Due to the number of incidents, incident outcomes and incident outcomes cases considered,
this study level provides the highest resolution.
Representative
Set
Bounding
Group
no single technique whose application guarantees the comprehensive listing of all inci
dents (i.e., the reality list of Figure 1.8 is unattainable). Nonetheless, use of hazard
identification techniques and Table 1.2 can lead to the identification of a broad spec
trum of incidents, sufficient for defining even the expansive list of incidents (Section
1.4.2.1).
Other approaches for enumeration of major incidents and their initiating events
have been developed. One of these uses fault tree analysis (FTA). The fault tree is a
logic diagram showing how initiating events, at the bottom of the tree, through a
sequence of intermediate events, can lead to a top event. This analysis requires two
knowledge bases: (1) a listing of major subevents which contribute to a top event of
loss of containment, and (2) the development of each subevent to a level sufficient to
describe the majority of initiating events. For enumeration, this process is executed
without any attempt to quantify the frequency of the top event. However, this fault
tree can serve as a means for obtaining frequencies later in the CPQRA. The success of
this technique is principally dependent on the expertise of the analyst. An example is
given by Prugh (1980).
The "Loss of Containment Checklist53 included in this book as Appendix A can be
applied to enumerate credible incidents. This checklist considers causes arising from
nonroutine process venting, deterioration and modification, external events, and pro
cess deviations. Sample incidents include the following:
• overpressuring a process or storage vessel due to loss of control of reactive mate
rials or external heat input
• overfilling of a vessel or knockout drum
• opening of a maintenance connection during operation
• major leak at pump seals, valve stem packings, flange gaskets, etc.
• excess vapor flow into a vent or vapor disposal system
• tube rupture in a heat exchanger
• fracture of a process vessel causing sudden release of the vessel contents
• line rupture in a process piping system
• failure of a vessel nozzle
• breaking off of a smallbore pipe such as an instrument connection or branch line
• inadvertently leaving a drain or vent valve open.
The reader should note, however, that the loss of containment checklist should not
be considered exhaustive, and other enumeration techniques should be considered in
developing an expansive list of incidents.
Another way to generate an incident list is to consider potential leaks and major
releases from fractures of all process pipelines and vessels. The enumeration of inci
dents from these sources is made easier by compiling pertinent information (listed
below), relevant to all process and storage vessels. This compilation should include all
pipework and vessels in direct communication, as these may share a significant inven
tory that cannot be isolated in an emergency.
This approach is discussed in more detail in the Rijnmond Area Risk Study
(Rijnmond Public Authority, 1982) and the Manual of Industrial Hazard Assessment
Techniques (World Bank, 1985). Of necessity, this approach excludes specific incidents
and initiating events that would be generated by hazard identification methods (e.g.,
releases from emergency vents or relief devices). Freeman et al. (1986) describe a
system that addresses both fractures and other initiating events. The list of incidents
can also be expanded by considering each of the incident outcomes presented in Table
1.2 and proposing credible incidents that can produce them. Pool fires might result
from releases to tank dikes or process drainage areas; vapor cloud explosions, flash
fires, and dispersion incidents from other release scenarios; confined explosions (e.g.,
those due to polymerization, detonation, overheating) from reaction chemistry and
abnormal process conditions; or BLEVE, from fire exposure to vessels containing
liquids.
1.4.2. Selection
The goal of selection is to limit the total number of incident outcome cases to be stud
ied to a manageable size, without introducing bias or losing resolution through over
looking significant incidents or incident outcomes. Different techniques are used to
select incidents (Section 1.4.2.1), incident outcomes (Section 1.4.2.2), and incident
outcome cases (Section 1.4.2.3). The risk analyst must be proficient in each of these
techniques if a defensible basis for a representative CPQRA is to be developed.
1.4.2.1. INCIDENTS
The purpose of incident selection is to construct an appropriate set of incidents for the
study from the initial list that has been generated by the enumeration process. An
appropriate set of incidents is the minimum number of incidents needed to satisfy the
requirements of the study and adequately represent the spectrum of incidents enumer
ated, considering budget constraints and schedule.
The effects of selection are shown graphically in Figure 1.8. The reality list con
tains all possible incidents. It approaches infinitely long. The initial list contains all the
incidents identified by the enumeration methods chosen. The remaining lists are
described in this section. Figure 1.8. shows the relative reductions in list size that are
achieved by successive operations on the initial list.
One of the risk analyst's jobs is to select a subset of the Initial List for further analy
sis. This involves several tasks, each resulting in a unique list ( Figure 1.8). Throughout
the selection process, the risk analyst must exercise caution so that critical incidents,
which might substantially affect the risk estimate, are not overlooked or excluded from
the study. The initial list of incidents is reviewed to identify those incidents that are too
small to be of concern (Step 4, Figure 1.4). Removing these incidents from the initial
list produces a revised list (Figure 1.8).
To be cost effective and reduce the CPQRA calculational burden, it is essential to
compress this revised list by combining redundant or very similar incidents. This new
list is termed the condensed list (Figure 1.8). This list can and should be reduced fur
ther by grouping similar incidents into subsets, and, where possible, replacing each
subset with a single equivalent incident. This grouping and replacement can be accom
plished by consideration of similar inventories, compositions, discharge rates, and dis
charge locations.
The list formed in this manner is the expansive list and represents the list from
which the study group is selected. A detailed or complex study would utilize the entire
expansive list of incidents, while a screening study would utilize only one or two inci
dents from this list.
The expansive list can be reduced to one or both of two smaller "lists55: the bound
ing group or the representative set (Section 1.3.1; and Figure 1.5). Selection of a
bounding group of incidents typically considers only the subsets of catastrophic inci
dents on the expansive list. This may be further reduced by selecting only the worst
possible incident or worst credible incident.
Selection of a representative set of incidents from the expansive list should include
contributions from each class of incident, as defined in Table 1.3. This process can be
facilitated through the use of ranking techniques. By allocating incidents into the three
classes presented in Table 1.3, an inherent ranking is achieved. Further ranking of indi
vidual incidents within each incident class is possible. Various schemes can be devised
to rank incidents within each incident class (e.g., preliminary ranking criteria based on
the severity of hazard posed by released chemicals, release rate, and total quantity
released). A ranking procedure is important in the selection of a representative set of
incidents if the study is to minimize bias or loss of resolution.
Ranking can also be a useful tool if the study objectives (Section 1.9.2) exclude
incidents below a specified cutoff value. One example is the establishment of a cutoff
for loss of containment of material events by specifying a limited range of hole sizes for
a wide range of process equipment (e.g., two for process pipework, one representing a
fullbore rupture and the other 10% of a full bore rupture). This approach is presented
in the Manual of Industrial Hazard Assessment Techniques (World Bank, 1985). Such a
cutoff is arbitrary and a more fundamental approach is to identify, from consequence
techniques (Chapter 2), the minimum incident size of importance for each of the mate
rials used onsite. This ensures consistent treatment of materials of different hazards.
Figure 1.9 (Hawksley, 1984) contains data on pipeline failures including the frequency
distributions for holes of various sizes.
1.4.2.2 INCIDENT OUTCOMES
The purpose of incident outcome selection is to develop a set of incident outcomes that
must be studied for each incident included in the finalized incident study list (i.e., the
bounding group, representative set, or expansive list of incidents). Each incident needs
to be considered separately. Using the list of incident outcomes presented in Table 1.2,
the risk analyst needs to deter nine which may result from each incident. This process is
not necessarily straightforward. While the analyst can decide whether an incident
LEGEND
Weep
PIPE FAILURES PER FOOT YEAR
Canadian Atomic
Energy
GuIfOiI
Vapor Cloud
Travels Downwind Pool Fire
(if not ignited) Occurs
No Ignition - No Ignition -
Toxic Vapor Toxic Vapor
Exposure Exposure
Pool Fire
Occurs
FIGURE ]. 10. Typical spill event tree showing potential incident outcomes for a hazardous
chemical release.
FIGURE 1.11. Spill event tree for a flammable gas release.
also the analyst's responsibility to recognize the sensitivity of the cost of the CPQRA to
each parameter and avoid wasting resources.
One effective strategy is to screen the parameter value ranges and select a minimal
number of outcome cases to complete a first pass risk estimate. Using sensitivity meth
ods, the importance of each selected parameter value can be determined, and adjust
ments made in subsequent passes, maintaining control of the growth of the number of
incident outcome cases while observing impacts on resulting estimates.
It is also useful to determine upper and lower bounds for the risk estimate using
the parametervalue range available. This offers the analyst a reference scale against
which to view any single point estimate, along with its sensitivity to changes in any
given parameter. Various mathematical models are available for determining the upper
and lower bounds for the parametervalue ranges available. These include techniques
commonly used in the statistical design of experiments (e.g., see Box and Hunter,
1961; Kilgo, 1988). These methods can be used to identify critical parameters from all
of the parameters identified. Linear programming techniques and min/max search
strategies (e.g., see Carpenter and Sweeny, 1965; Long, 1969; Nelder and Mead,
1964; Spendley et al, 1962) can be used thereafter to find values for these critical
parameters that will produce both the upper and lower bounds (maximum and mini
mum values) for the risk estimate.
I U the ReleaM Ia There Immediate1 Does a Pool Does the Pool
I Instantaneous? Ignition? Form? Ignite?
FIGURE 1.12. Spill event tree for a flammable liquid release.
1.4.3. Tracking
FIGURE 1.12. Spill event tree for a flammable liquid release.
1.4.3. Tracking
1.5.1. Screening Techniques
In creating a screening program, it is helpful to determine the organizational levels that
are most amenable to screening, and those where CPQRAs can be applied most effec
tively. Figure 1.13 illustrates the structure of a typical CPI organization. It shows a
hierarchical scheme, with the organization divided into facilities (plants), the facilities
divided into process units, the process units divided into process systems and the pro
cess systems divided into pieces of equipment. A general observation is that the
number of possible CPQRAs increases exponentially—but that the scope of each one
narrows—moving from the top to the bottom of the hierarchy. Use of CPQRA is typi
cally restricted to the lower levels of the hierarchy, and in those levels it is selectively
applied.
Methods are needed to screen—prioritize and select—process units, systems, and
equipment for selective application of CPQRA. These methods must ensure that all
facilities are considered uniformly in the screening process.
Establishment of a prioritized listing of candidate studies allows efforts to focus on
the most onerous hazards first and, depending on available resources, progress to less
serious hazards. Certain listings are "zoned35 according to high, medium, and low levels
of concerns, and studies placed into the lowest class receive attention only after all stud
ies in higher classes have been executed. If a decision is made to zone a priority list, it is
important to establish zone cutoff criteria prior to screening in order to avoid bias.
Bask estimates can be developed at any level of the typical CPI organization, but
usually focus on specific elements of the lower levels of the hierarchy—for instance, the
COMPANY
HEADQUARTERS
MANUFACTURING
FACILITIES
PROCESS
UNITS
PROCESS
SYSTEMS
PROCESS
EQUIPMENT
FIGURE 1.13. Structure of a typical CPI company.
risk from the rupture of a storage tank. The following discussions of screening methods
show that methods are available to study various levels of the typical CPI organization.
1.5.1.2. INVENTORYSTUDIES
The inventories of hazardous materials should be itemized (including material in pro
cess, in storage, and in transport containers). The information should include signifi
cant properties of the material (e.g., toxicity, flammability, explosivity, volatility),
normal inventory and maximum potential quantity, and operating or storage condi
tions. In some cases, screening can, or must, be done by means of government specifi
cations (New Jersey, 1988, and EEC's "Seveso Directive,5' 1982).
Major hazards can be identified from an inventory study. Where these are toxic
hazards, simple dispersion modeling—assuming the worst case and pessimistic atmo
spheric conditions—can be performed. Where fires or explosions are the hazards, simi
lar simple consequence studies may be made. Estimated effect zones can be plotted on a
map to determine potential vulnerabilities (population at risk, financial exposure, busi
ness interruption, etc.); for screening purposes, estimates of local populations may be
sufficient. Of course, when significant vulnerabilities are found, more thorough studies
may be required.
1.5.1.3. CHEMICALSCORING
Various systems have been developed to assign a numeric value to hazardous chemicals
using thermophysical, environmental, toxicological, and reactivity characteristics. The
purpose of each system is to provide an objective means of rating and ranking chemi
cals according to a degree of hazard reference scale. Three of these methodologies are
systems proposed by the NFPA 325M (1984), the U.S. EPA (1980, 1981), and
Rosenblum et al. (1983).
NFPA has a rating scheme that assigns numeric ratings, from O to 4, to process
chemicals. These ratings represent increasing health, flammability, and reactivity haz
ards; the fourth rating uses special symbols to denote special hazards (e.g., reactivity
with water). This system is intended to show firefighters the precautions that they
should take in fighting fires involving specific materials; however, it can be used as a
preliminary guide to process hazards. The U.S. EPA has developed methods for rank
ing chemicals based on numerical values that reflect the physical and health hazards of
the substances. Rosenblum et al. (1983) give an index system that assigns numerical
values to the various hazards that chemicals possess and that can be used to prioritize a
list of chemicals. This technique is more complex and lesspracticed than the NFPA
diamond system.
1.5.3. Applications within New Projects
The depth of study presented in Table 1.6 directly applies to new projects as well. The
main distinction between new projects and existing facilities (Figure 1.7) is the infor
mation available for use in the CPQRA. Early in a new project, information is con
strained, limiting the depth of the study. This constraint is virtually nonexistent for
existing facilities. As a new project progresses, the information constraint is gradually
removed.
TABLE 1 .6. Applicability and Sequence Order of Depth of Study for Existing Facilities
Risk estimation technique*
Organizational
hierarchy level Depth of study Consequence Frequency Risk
Process unit Simple/consequence 1 2 3
Intermediate/frequency 1 2 N.A.
Complex/risk 1 N.A. N.A.
Process system Simple/consequence 1 2 3
Intermediate/frequency 1 2 3
Complex/risk 1 2 3
Equipment Simple/consequence 1 2 3
Intermediate/frequency 1 2 3
Complex/risk 1 2 3
4
NA, not applicable; 1, First task in series; 2, second task in series; 3, third task in series.
TECHNICAL
MANAGEMENT
CUMULATIVE FREQUENCY
FIGURE 1.14. Frequency distribution of the failure rate ratio collected by the National Centre
of Systems Reliability over the period 19721987 From Ballard (1987), reprinted with
permission.
ANNUAL SYSTEM FAILURE PROBABILITY
Base. The teams were also allowed complete freedom in making assumptions, selecting
incidents to study, choosing failure rate data, etc. Figure 1.15 shows that the resulting
estimates ranged over several orders of magnitude, well beyond the range of uncer
tainty calculated by some of the teams. When the teams were subsequently directed to
follow similar assumptions, the resulting estimates converged to a much more accept
able range (i.e., within a factor of 5). This study and its implications is discussed in
more detail in Chapter 4. Consequently, it is important to recognize that along with
the technical uncertainties associated with models and data discussed elsewhere in this
book, the essence of the accuracy and corresponding uncertainty of a risk estimate also
depend heavily upon the expertise and judgment of the analyst. The need to document
and review such assumptions is discussed in depth in Section 1.9.5.3 on Quality
Assurance.
Existing techniques
Codes and standards 100
Unstructured hazard identification (e.g., indices, judgment) 95
Structured hazard identification (e.g., HAZOP, FMEA) 60
CPQRA techniques
Consequence estimation 40
Frequency estimation 30
Risk estimation 20
Use of risk targets 10
This section offers an overview of the role of CPQRA project management. A CPQRA
must be carefully managed in order to obtain the required results in a timely and cost
effective manner. Project management tasks include study goals (Section 1.9.1), study
Frequency of N or More Fatalities Per Year
Number of Fatalities Per Event
FIGURE 1.16. Acceptable risk criteria. AL>\RA, as low as reasonably achievable.
1.9.1. Study Goals
Section 1.3.2 and Table 1.4 describe typical study goals. These can originate from
external sources, such as regulatory agencies, or from internal initiatives (e.g., senior
management).
1.9.2. Study Objectives
It is critical for project management to understand the study goals and to firmly estab
lish study objectives. The study objectives define the project goals in precise terms that
USER DEFINE GOALS OF CPQRA
REQUIREMENTS (TABLE 1.4)
(§1.9-1)
DEFINE DOCUMENTATION
REQUIREMENT TO
SATISFY USER
(§1.9.4)
STUDY ACCEPTED
FIGURE 1.17. Logic diagram for CPQRA project management.
1.9.3. Depth of Study
A careful determination of the depth of study is essential if CPQRA goals and objec
tives are to be achieved, adequate resources are to be assigned, and budget and sched
ules are to be controlled. The calculation workload for a given depth of study can
expand factorially as one moves from the origin along any one of the axes of the study
cube (Section 1.3.1). It is essential to estimate this calculation burden prior to finaliz
ing a depth of study so that project costs and schedule requirements can be evaluated. A
risk analyst and a risk methods development specialist can provide project management
with valuable assistance in estimating this workload and with guidance in selecting an
appropriate depth of study.
Figure 1.18 presents a schematic for determining the appropriate depth of study.
Basically, given an approved scope of work, which specifies the risk measures to be cal
culated and presentation formats to be used, the analyst needs to select the following
(Section 1.3.1):
• the appropriate risk estimation technique
• the appropriate complexity of study
• the appropriate number of incidents.
Once values have been assigned to each of these study parameters, the depth of
study—cell within the study cube given in Figure 1.5—has been determined.
SELECT APPROPRIATE
COMPLEXITY OF STUDY
SELECT APPROPRIATE
NUMBER OF INCIDENTS
"Parameters listed may or may not apply in the following formula to estimate the study's calculation burden:
Number of calculations = |jX,
i=\
where n = number of applicable parameters and X1 = study parameters from above listing.
1.9.5. Construction of a Project Plan
A written project plan should be prepared for every CPQRA, regardless of the scope of
work or depth of study. The circulation and availability of such a plan to members of
the project team provides for communication, team building, and direction. It is only
through the preparation of such a written plan that aspects of the study critical to its
success receive adequate attention.
Various texts on project management offer useful guidance on preparing a project
plan, including suggested plan contents. This material need not be presented here.
However, there are aspects of a project plan for a CPQBA that are unique and these are
discussed in the following sections.
HAZARD IDENTIFICATION & INCIDENT SELECTION
SIMPLE/ INTERMEDIATE/ COMPLEX/
DEPTH OF STUDY CONSEQUENCE FREQUENCY RISK
PEOPLE PE PE PEfRA
EFFORT 0.51. 8 MW 12MW 24MW
TOOLS What if/PHA Course HAZOP HAZOP/FMEA
DATA Concept PFO PFD.P&ID
(Export shown
layout)
INCIDENTS 26 1820 90100
IDENTIFIED
PROGRESSION THROUGH DEPTHS OF RISK ESTIMATION
STUDY (REFER TO FIGURE 1.4) SIMPLE/ INTERMEDIATE/ COMPLEX/
DEPTH OF STUDY CONSEQUENCE FREQUENCY RISK
FINDINGSOF INITIALCPQRA REQUIRE PEOPLE PE MW PE+RA RA
INCREASED DEPTH OF STUDY EFFORT 00.05 0.051 MW 25MW
TOOLS MINIMAL MINIMAL OR SIMPLE OR
MODELS SIMPLE DETAILED
COMPUTER COMPUTER
PACKAGE PACKAGE
ties have not been included in the totals presented. Administration of the project may
require an additional 510% of the total manpower estimates presented.
1.9.5.2. PROJECT SCHEDULING
Table 1.10 provides guidance on the total manpower required for a risk analysis. The
elapsed time is a function, to some degree, of the number of personnel provided, but
there is an inherent task structure in each depth of study that constrains project man
agement from paralleling all individual tasks. Consequence and frequency analyses can
be done in parallel, but must logically follow hazard identification and incident selec
tion. Final risk estimation must await completion of the consequence and frequency
analyses.
TABLE 1.10. Manpower Requirements for Depths of Study of a Single Process
System (UNIT)
"Note that the data presented have units of personweeks. These data also need to be converted to calendar weeks
by the project manager through development of a project schedule. The resulting number of calendar weeks may
be substantially greater than the values shown above, depending on availability of critical personnel, tools, train
ing opportunities, etc.
6
ThC values presented do not include project management activities, which can be estimated as an additional
burden of 510% of the totals shown. Sensitivity studies are also not included and are often required to evaluate
potential risk mitigation measures.
1.9.5.3. QUALITYASSURANCE
The first step in quality assurance is to ensure the adequacy and availability of staff and
resources for the study. Since CPQRA is a relatively new CPI technology, it is likely
that the expertise of staff support will be deficient in certain technology areas. Conse
quently, quality assurance is a critical check and balance procedure of any CPQBA pro
ject plan. Adequate resources need to be assigned to quality assurance as a line item in
the project plan.
Early risk analysis studies (e.g., Rijnmond Public Authority, 1982) were routinely
passed on to independent reviewers. These reviews were budgeted at up to 10% of the
primary budget. Such outside reviews are now less common, but are appropriate for
organizations relatively inexperienced in CPQRA. Alternatively, outside experts may
be commissioned to undertake the study. Their activities can be monitored by com
pany staff. This monitoring may be done by periodic meetings or by a staff member
assigned to the review team.
Such peer reviews or reviews by corporate staff of outsideexpert work products
are only one of several layers of reviews that can be built into the project plan to ensure
TABLE 1.11. CPQRA Reviews and Purposes
Plant staff
Reveal any misrepresentation of plant practices, existing hardware and process configurations, facility
operational data, and site characteristics
Corporate staff
Ensure consistency with previous CPQRA formats, adherence to company CPQKA practices,
adequacy of documentation, etc. If staff includes risk analysts, provide peer review functions to the
project team
Peer or expert review
Review should be carried out by competent risk analysts not involved in the CPQRA. Review should
focus on appropriateness of methods, quality and integrity of the data base used, validity and
reasonableness of assumptions and judgments, as well as recommendations for further study
Management
Assuming the role of user, management should be satisfied that the report meets its requirements
completely, in line with the agreed on scope of work and that all conclusions and recommendations, if
any, are thoroughly understood
Data compilation
• Data, should, be checked as being correct, relevant and uptodate
• Data on chemical toxicity should be reviewed for reasonableness
• Documentation of the sources of data used should be maintained
Consequence estimation
• Models should be well documented
• Trial runs should be compared against known results for validation (to protect against misunderstanding of
model requirements)
• Consequence results should consider all important effects (e.g., explosion analysis should include blast and
thermal radiation effects)
• Effect models should correspond to the study objectives
• Documentation of input data and results should be maintained
Frequency estimation
• Historical data should be confirmed as being applicable
• Fault and event tree model results should be confirmed against the historical record where feasible
• Documentation of the frequency estimation should be maintained
Risk estimation
• Results should be checked against experience for reasonableness
• Audit trail of documentation should he maintained
It is important to note that wellconstructed and wellexecuted CPQRAs rely
heavily on judgment. Short training programs provide users with the necessary tools;
however, judgment can come only from the experience of applying them. Project man
agement must be aware that estimates from inexperienced practitioners need greater
scrutiny than those from accomplished risk analysts.
1.9.5.5. PROJECT COST CONTROL
As CPQBAs can consume substantial resources, attention to cost control in developing
a project plan is essential. Once funding has been approved, it is important to docu
ment the allocation of that funding to accomplish the study. This allocation covers
• manpower costs (internal to the organization)*
• tool acquisition and installation (hardware and software) *
• data acquisition* computer costs
• training costs
• travel
• publication and presentation
• outside consultant services (all types)*
• project overheads
The four starred (*) items above offer unique problems for CPQBA project man
agers. They represent greater uncertainty in preparing project cost estimates than do
the other contributors. Consequently, greater effort to define them for estimate pur
poses is required, and greater attention to them through cost control procedures
during the project is necessary. The project manager must rely on the risk analyst for
estimating model development costs, software acquisition costs, outside consulting
services, and data acquisition expense.
Because of the potential for uncertainty, it is good practice to require that the risk
analyst provide documentation for cost estimates, including statements from any antic
ipated source of outside service (e.g., consultants, data acquisition). For example, if the
scope of work required earthquake analysis and this was beyond the capabilities of the
organization's staff, it would be necessary to provide at least preliminary estimates for
this analysis from outside firms. While this may require additional effort in preparing
resource requirements, this effort should result in better definition of costs prior to
project approvals and the avoidance of cost overruns thereafter. Such documentation
can also be used as input to cost control procedures over the life of the project. Other
wise, routine project cost controls in use for managing capital projects can be applied.
1.9.6. Project Execution
• Technology
Process Safety Information. The CPQRA provides a current summary of
hazards on the site and a listing and summary of all important relevant docu
ments.
Process Risk Analysis. This is the primary function of the CPQRA, one that
must be kept up to date and made available to new staff.
Management of Change (Technology). All changes/modifications should be
subjected to the same rigor of analysis as the original study.
Rules and Procedures. These should be developed in the context of the
CPQRA results.
• Personnel
Staff Training. The CPQRA presents insights to specific facility risks with all
relevant documents appended or referenced.
Incident Investigation. The CPQRA can be useful in incident investigation,
to check whether the event was properly identified and if protective systems
performed as expected. If not previously identified, it should be added to the
CPQEA and the results recalculated. Additional risk reduction measures may
be suggested.
Auditing. The CPQRA can serve as a guide to the auditor to familiarize the
auditor with major risk contributors and past studies of them.
• Facilities
Equipment Tests and Inspections. The CPQEA highlights the importance
of testing intervals in maintaining protective system reliability. Regular checks
are necessary to ensure these are maintained.
Prestartup Safety Reviews. This function is similar to the auditing role.
Important features are identified for inspection and checking.
Management of Change (Facilities). See Management of Change (Tech
nology).
• Emergencies
CPQEAs can assist in developing a site emergency response plan.
• Some Additional Uses (not specific to the site risk management program)
Community Relations. Discussions with the local community are often aided
by the availability of uptodate CPQRAs.
Plant Comparisons. Many companies operate several plants of similar design.
CPQRA data from one can be used as a guide for new plants or for modifying
other existing plants.
Operating Standards. All the CPQRA component techniques make assump
tions of how the plants should be operated (HAZOP, fault tree failure frequen
cies, consequence calculations, etc.). When documented and kept current, these
can be checked at a later stage for accuracy.
It is important to recognize that a CPQRA shows whether a plant can operate at a
given risk level, but cannot ensure that the plant will operate consistent with the
assumptions used to estimate risk. Naturally, if actual operations differ from study
assumptions, the risk estimates produced cannot be considered representative. Study
assumptions need to reflect reality, and as reality changes, so must study assumptions.
Corresponding risk estimates will need to he undated. Updates can be triggered by
• process changes (e.g., hardware, software, material, procedures), availability of
improved input data (e.g., toxicology data)
• introduction of company risk targets
• advances in CPQRA component techniques
• changes to company property (e.g., neighboring process units, administration
building relocation)
• changes in neighboring property (e.g., expansion of a housing development to
company property limits)
Maintenance of a CPQRA means much more than assuming the availability of a
copy of the original study in an organization's files, though it is important to preserve
and store the results in a secured system. The need to maintain the study should be rec
ognized and accepted at the time the commitment is made to execute the CPQRA. As
with any process documentation, without such commitment, the CPQRA report will
gradually hut assuredly become dated and lose its value to the company's risk manage
ment program.
1,11. References
AIChE/CCPS (1992), Guidelines for Hazard Evaluation Procedures second edition with worked
examples. Center for Chemical Process Safety, American Institute of Chemical Engineers,
New York.
AIChE/CCPS (1995), Tools for Making Acute Risk Decisions, Center for Chemical Process Safety,
American Institute of Chemical Engineers, New York.
AIChE/CCPS (1995), Guidelines for Process Safety Documentation, Center for Chemical Process
Safety, American Institute of Chemical Engineers, New York.
AIChE/CCPS (1994), Guidelines for Implementing Process Safety Management Systems, Center for
Chemical Process Safety, American Institute of Chemical Engineers, New York.
AIChE/CCPS (1989), Guidelines for Technical Management of Chemical Process Safety, Center for
Chemical Process Safety, American Institute of Chemical Engineers, New York.
AIChE/CCPS (1995), Plant Guidelines for Technical Management of Chemical Process Safety,
Center for Chemical Process Safety, American Institute of Chemical Engineers, New York.
AIChE/CCPS (1988a). Guidelines for Safe Storage and Handling of High Toxic Hazard Materials.
Center for Chemical Process Safety. American Institute of Chemical Engineers, New York.
Amendola, A. (1986). "Uncertainties in Systems Reliability Modeling: Insight Gained Through
European Benchmark Exercises," Nuclear Engineering Design 93, 215225, Amsterdam,
The Netherlands: Elsevier Science Publishers.
Arendt, J. S. et al. (1989). ^4 Manager^ Guide to Quantitative Risk Assessment of Chemical Process
Facilities. JBF Associates, Inc., Knoxville, Term., Report No. JBFA11988, prepared for the
Chemical Manufacturers Association, Washington, D.C.; January.
Ballard, G. M. ( I 987). "Reliability Analysis—Present Capability and Future Developments."
SRS Quarterly Digest System Reliability Service, UK Atomic Energy Authority,
Warrington, England, pp. 311, October.
Batsone, R. J. (1987). Proceedings of the International Symposium on Preventing Major Chemical
Accidents. Washington, D.C. (J. L. Woodward, ed.). American Institute of Chemical Engi
neers, New York. Feb. 35.
Box. G. E. P. and Hunter, J. S. (1961). "The 2kp Fractional Factorial Designs. Part 1,"
Technometrics 3(3), 311346.
Boyen, V. E. et al. (1988). "Process Hazards Management." Document developed by Organiza
tion Resource Counselors, Inc. (ORC) [submitted to OSHA for future rulemaking on pro
cess hazards management], Washington, D.C.
Bretherick, L. (1983). Handbook of Reactive Chemical Hazards, 2nd edition. London:
Butterworths.
Carpenter, B. H., and Sweeny, H. C. (1965). "Process Improvement with 'Simplex3
SelfDirecting Evolutionary Operation." Chemical Engineering 72(14), 117126,
DeHart, R., and Gaines, N. ( I 987): "Episodic Risk Management at Union Carbide." AIChE
Spring National Meeting, Symposium on Chemical Risk Analysis, Houston. American
Institute of Chemical Engineers. New York.
Dow3s Fire and Explosion Index—Hazard Classification Guide, 7th edition, 1994. CEP Technical
Manual, American Institute of Chemical Engineers, New York.
Daw's Chemical Exposure Index Guide, 1994 . CEP Technical Manual, American Institute of
Chemical Engineers, New York.
EEC (19X2). "MajorAccident Hazards of Industrial Activities" ("Seveso Directive"). European
Economic Community. Council Directive X2501EEC Official Journal (OJ) Reference
No. L23), 5.8.1982: Amended October 1982. [Available from European Economic Com
munity, Press and Information Services, Delegation of the Commission of European Com
munities, Suite 707, 210OM Street, N.W., Washington, D.C.]
Freeman, R. A. (19X3). "Problems with Risk Analysis in the Chemical Industry." Plant/Opera
tions Progress 2(3), 185190.
Freeman. R. A. et al. (1986). "Assessment of Risks from Acute Hazards at Monsanto." 1986
Annual Meeting, Society for Risk Analysis. Nov. 912, Boston, MA. Society for Risk Analy
sis. 8000 West Park Drive, Suite 400, McLean. VA 2210'.
Gibson, S. B. (1980). "Hazard Analysis and Risk Criteria." Chemical Engineering Progress
(November), 4650.
Goyal, R. K. (19X5). "PRATwo Case Studies from the Oil Industry." Paper presented at Ses
sion 5A of Reliability C85, Symposium Proceedings, July 1012, 1985; VoI 2, p. 5A/3.
Jointly sponsored by National Centre of Systems Reliability. Warrington, England and
Institute of Quality Assurance, London, England.
Hawksley, J. L. (1984). Some Social, Technical and Economical Aspects of the Risks of Large Chemical
Plants. Chemrawn III, World Conference on Resource Material Conversion, The Hague,
June 2529.
Health & Safety Executive (1978). Canvey—An Investigation of Potential Hazards from the Opera
tions in the Canvey Island/Thurrock Area. 195 pp. HMSO. London, UK.
Health & Safety Executive (1981). Canvey—A Second Report, 130 pp. HMSO, London, UK.
Helmers, E. N and L. C. Schaller (1982). "Calculated Process Risk and Hazards Management."
AIChE Meeting, Orlando, FL, Feb. 20Mar. 3. American Institute of Chemical Engineers,
New York.
IChemE (1985). Nomenclature of Hazard and Risk Assessment in the Process Industries. Institution
of Chemical Engineers, UK.
Hendershot, D. C. (1996) "Risk Guidelines as a Risk Management Tool," 1996 Process Plant
Safety Symposium, Houston, TX, April 12, 1996, Session 3
ICI (Imperial Chemical Industries) (1985). TheMondIndex, 2nd edition. ICI PLC, Explosion
Hazards Section. Technical Department. Winnington, Northwick, Cheshire CW8 4DJ,
England.
Joschek K. T. (1983). "Risk Assessment in the Chemical Industry." Plant/Operations Progress
2(1 January), 15.
Kaplan, S., and B. J. Garrick ( 1981). "On the Quantitative Definition of Risk." Risk Analysis
1(1), 1127.
Kilgo, M. B. (1988). "An Application of Fractional Factorial Experimental Designs." Quality
Engineering, \, 1923. American Society for Quality Control and Marcel Dekker, New
York.
Lees, F. P. (1980). Loss Prevention in the Process Industries, 2 Volumes. Butterworths. London
and Boston.
Long, D. E. (1969). "Simplex Optimization of the Response from Chemical Sys terns. "Anal.
Chim. Acta 46,193206.
Marshall, V. C. (1987). Major Chemical Hazards. Wiley, New York.
Mudan, K. S. (1987). "Hazard Ranking for Chemical Processing Facilities." ASME Winter
Annual Meeting, Boston, MA. Dec. 1318. American Society of Mechanical Engineers,
New York.
Nelder, J. A., and Mead, R. (1964). "A Simplex Method for Function Minimization." The Com
puter Journal 7, 308313.
Mundt, A. G. (1995). "Process Risk Management for Facilities and Distribution." AIChE
Summer National Meeting, Boston, July 30Aug 2, American Institute of Chemical Engi
neers, New York.
New Jersey (1983). 'Toxic Catastrophe Prevention Act Program." State of New Jersey,
N.J.A.C. 7: 1, 2, 3, 4 and 6. New Jersey Register, Monday, June 20, 1988, 20 N.J.R. 1402.
NFPA 325M (1984). Fire Hazard Properties of Flammable Liquids, Gases, and Volatile Solids.
National Fire Protection Association, Quincy, MA 02269.
NUREG (1983). PRA Procedures Guide: A Guide to the Performance of Probabilistic Risk Assessment
for Nuclear Power Plants, 2 volumes, NUREG/CR2300, U.S. Nuclear Regulatory Commis
sion, Washington, D.C. (available from NTIS).
NUREG (1984). PZM Status Review in the Nuclear Indusrry, NUREG1050, Nuclear Regula
tory Commission, Washington D.C. September, 1984 (available from NTIS).
NUREG (1985). Probabilistic Safety Analysis Procedures Guide, NUREG/CR2815, Nuclear
Regulatory Commission, Washington D.C. August, 1985 (available from NTIS).
Ormsby, R. W. (1982). "Process Hazards Control at Air Products." Plant/Operations Progress 1,
141144.
Pikaar, J. (1995). "Risk Assessment and Consequence Models." 8th International Symposium on
Loss Prevention and Safety Promotion in the Process Industries^ June 69, 1995, Antwerp, Bel
gium, Keynote lecture on Theme 4.
Pikaar, M. J., and M. A. Seaman (1995). A Review of Risk Control. Zoetermeer, Netherlands:
Ministrie VROM.
PiIz, V. (1980). "What Is Wrong with Risk Analysis?" 3rd International Symposium on Loss
Prevention and Safety Promotion in the Process Industries, Basle, Switzerland, 6/448454.
Swiss Society of Chemical Industries, September 1519.
Poulson, J.M. et al. (1991). "Managing Episodic Incident Risk in a Large Corporation." AIChE
Summer National Meeting, Pittsburgh, Aug 1821, American Institute of Chemical Engi
neers.
Prugh. R. W. (1980). "Application of Fault Tree Analysis." Chemical Engineering Progress July,
5967.
Rijnmond Public Authority (1982). A Risk Analysis of 6 Potentially Hazardous Industrial Objects
in the Rijnmond Area—A Pilot Study. D. Reidel, Dordrecht, the Netherlands and Boston,
MA.
Renshaw, P.M. (1990). "A Major Accident Prevention Program." Plant/Operations Progress
9(3), 194197.
Rosenblum, G. R. et al. (1983). "Integrated Risk Index Systems." Proceedings of the Society for
Risk Analysis. Plenum Press, New York, 1985.
Seaman, M. A., and Pikaar, M. J., "A Review of Risk Control," VROM, 11030/150, June, 1995.
Spendley. W. et al. (1962). "Sequential Application of Simplex Designs in Optimisation and
Evolutionary Operation." TechnomePncs 4(4), November.
TNO (1979). Methods for the Calculation of the Physical Effects of the Escape of Dangerous Materials:
Liquids and Gases, 2 Volumes. P.O. Box 312, 7300 AH Apeldoorn, The Netherlands.
Tyler, B. J. et al. (1996), "A toxicity hazard index," Chemical Health & Safety., January/February,
1996,1925.
USCIP Working Party (1985). "Standard Plan for the Implementation of Hazard Studies 1:
Refineries." Union des Chambres Syndicales de 1' Industrie de Petrole (UCSIP), Paris,
France.
US EPA (1980). "Chemical Selection Method: An Annotated Bibliography": Toxic Integration
Information Series. EPA 560/TTIS80001, November (available from NTIS).
US EPA (1981). "Chemical Scoring System Development," by R. H. Ross and P. Lu, Oak
Ridge National Laboratory. Interagency Agreement No: 79Dx9856, June (available from
NTIS).
Van Kuijen, C. J. ( I 987). "Risk Management in the Netherlands: A Quantitative Approach."
UNIDO Workshop on Hazardous Waste Management and Industrial Safety, Vienna. June
2226.
Warren Centre (1986). Hazard Identification and Risk Control for the Chemical and Related Indus
tries—Major Industrial Hazards Project Report (D. H. Slater, E. R. Corran, and R. M.
Pithlado, eds.). University of Sydney, NSW 2006. Australia.
Watson, S. R. (1994). "The Meaning of Probability in Probabilistic Risk Assessment." Reliabil
ity Engineering and System Safety 45, 261269.
Watson, S. R. (1995). "Response to Yellman andMurray's comment on The meaning of proba
bility in probabilistic risk analysis'." Reliability Engineering and System Safety 49,207209.
World Bank (1985). Manual of* Industrial Hazard AssessmentTechniques. Office of Environmen
tal and Scientific Affairs. World Bank, Washington, D.C.
Yellman, T. W., and Murray, T. M. (1995). "Comment on The meaning of probability in
probabilistic risk analysis'." Reliability Engineering and System Safety 49, 201205.