Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 224–233

Contents lists available at ScienceDirect

Process Safety and Environmental Protection

journal homepage: www.elsevier.com/locate/psep

Conducting HAZOPs in continuous chemical processes:

Part II. A new model for estimating HAZOP time and a
standardized approach for examining nodes

Jordi Dunjó a,∗ , Vasilis M. Fthenakis b , R.M. Darbra a , Juan A. Vílchez a , Josep Arnaldos a
a Centre d’Estudis del Risc Tecnològic (CERTEC), Chemical Engineering Department, Universitat Politècnica de Catalunya, Diagonal 647,
08028 Barcelona, Catalonia, Spain
b Center for Life Cycle Analysis (CLCA), Department of Earth & Environmental Engineering, Columbia University, 500 West 120th St,

New York, NY 10027, USA

a b s t r a c t

The HAZOP organization phase entails two main tasks to ensure the success of the study, especially when reviewing
complex continuous chemical processes (e.g., petroleum-refining processes). The first task deals with selecting the
nodes, as we discussed in Part I of this paper. It addresses that task by proposing and justifying tools for and criteria
on how to break a process into manageable sections that could be reviewed independently (i.e., nodes selection).
Part II describes the development of a time-estimation model for planning HAZOP sessions. Its practical value was
confirmed with field work and data analyses of five HAZOPs. Furthermore, we focus on optimizing the time spent in
examining selected nodes. This papers also introduces a Deviations Structural Hierarchy (DSH) for treating deviations.
Finally, considering the Nodes Selection Methodology (NSM) defined in paper I, the HAZOP time-estimation model
(HTEM), and the Deviations Structural Hierarchy (DSH), we present the key tools, criteria, and guidelines for leading
HAZOPs for highly complex processes.
© 2011 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

Keywords: Process hazard analysis; HAZOP; HAZOP time-estimation model; Deviations Structural Hierarchy; Nodes
Selection Methodology

1. Introduction
One important factor to consider when managing HAZOP
studies is the time required to execute the entire analysis,
especially when numerous P&IDs must be reviewed. Through
our field work and data analysis described in Part I of this
paper, we developed a mathematical model to predict the
expected time to carry out a HAZOP study in continuous
chemical processes (e.g., petroleum-refining processes). As
explained therein, five HAZOPS were conducted following the
proposed criteria. The analysis of the findings proved valuable
in formulating an improved HAZOP time-estimation model,
thereby simplifying the current cumbersome ones. Two mod-
els were available prior to the present one. Thus, Freeman et al.
(1992) ma de the first attempt to plan HAZOP studies with an
expert system, establishing a way to assess how long and how
many work hours a HAZOP study entails. The authors based
their estimate on the number of major equipment items to be
analyzed, the system’s complexity, and the experience of the
HAZOP team leader. Five years later, Khan and Abbasi (1997)
improved this model, adding new factors and variables; viz.,
preparation time, meeting time, delays, and report writing; it
incorporates multivariable empirical equations. Additionally,
the preparation time and study time are the function of three
parameters: The number of P&IDs, the complexity of P&IDs,
and the skills of the team leader.
The present model aims to simplify and improve prior ones
by considering new predictors that define the complexity of
the process, and by avoiding subjective variables (i.e., the team
leader’s experience). Certainly, not account for the experience

∗
Corresponding author. Tel.: +34 669 53 99 81.
E-mail address: jdunjo@hotmail.com (J. Dunjó).
Received 12 August 2010; Received in revised form 16 February 2011; Accepted 1 March 2011
0957-5820/$ – see front matter © 2011 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.psep.2011.03.002
 Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 224–233 225

Table 1 – Comparison between HAZOP time estimation models.

Model Freeman et al. (1992) Khan and Abbasi (1997) Proposed model
TH = TP + TS + TW TH = TP + TS + TW + TD TH = TP + TS + TW

Preparation time
estimation (TP )
Description Tp depends on the number Tp depends on the number of Tp depends on the number of
of P&IDs and its complexity P&IDs and its complexity (simple, P&IDs (P&IDs), the total number of
(simple, standard, complex, standard, complex, very complex) pieces of major equipment (ME),
very complex) by counting by counting the number of pieces and requires to follow the
the number of pieces of of equipment, and pipelines per proposed Nodes Selection
equipment, pipelines and P&ID Methodology (NSM)
interlocks per P&ID
Input data Number of P&IDs, and Number of P&IDs, and P&IDs P&IDs, ME, NSM
P&IDs complexity complexity
Output data Tp Tp Tp ; and according to Part I: the
number of nodes (Nd)
Sessions Time
Estimation (TS )
Description TS depends on the leader TS depends on the leader skills TS depends on the number of
skills (novice, average or (novice, moderately experienced, nodes (Nd) = f(P&IDs, ME), and
experienced; according to experienced and highly requires to follow the proposed
the number of previous experienced; according to the Deviations Structural Hierarchy
HAZOPs carried out), the number of previous HAZOPs (DSH)
number of P&IDs, and carried out), the number of P&IDs,
P&IDs complexity and P&IDs complexity
Input data Number of P&ID, P&IDs Number of P&IDs, P&IDs P&IDs and ME (or which is the
complexity, leader skills complexity, leader skills same: Nd), DSH
Output data TS TS TS
Writing Time
Estimation (TW )
Description TW depends on the Tp TW depends on the Tp Tp depends on the Tp
Input data Tp Tp Tp
Output data TW TW TW
Delay time
estimation (TD )
Description Not considered TD includes the time lapsed due to Not considered
non-availability of members,
documents, or any other essential
items, and individuals responding
time
Input data Tp , TW
Output data TD

of the team leader will be possible only if we propose crite-
ria to guide leaders during the HAZOP performance. Thus,
our proposed model avoids including subjective variables, but
only if the following two approaches are adopted when the
organization- and execution-phases of the HAZOP analysis:
(1) the Nodes Selection Methodology (NSM) (explained in the
part I of the present paper) and (2) the Deviations Structural
Hierarchy (DSH), which will be illustrated after showing the
new model. Table 1 summarizes the key similarities and dif-
ferences between these models.
2. The HAZOP time-estimation model
During the analyses of the five HAZOPs, we collected and
recorded key variables for modeling purposes. Different
parameters were studied: (1) time to collect and organize the
key data needed for the study (TP ); (2) time to execute the
HAZOP sessions (TS ); and (3) time to prepare the first draft of
the HAZOP report (TW ). These three parameters were explored
to find relationships between factors that inherently define
the complexity of the process to be “HAZoped” (i.e., the num-
ber of pieces of major equipment, of P&IDs, and of PFDs,
the total amount of “minor” equipment, e.g., FCVs, Pumps
present in the process). After defining many combinations (all
studied both from the point of view of mathematics and of
process safety) we were able to establish a well-fitted regres-
sion between the time expected to complete a HAZOP study
in continuous chemical processes and the ambit of the var-
ious combinations. This modeling approach is equivalent to
that explained in part I of this paper when modeling the
expected number of nodes to be selected. The two predictors
enable us to easily to evaluate the complexity of the process
are, again, the following: (1) The number of pieces of major
equipment present in the processes (ME), which are clearly
illustrated on PFDs and (2) the number of P&IDs required to
define the process (P&IDs). Furthermore, we also recorded the
time required to brainstorm each selected node for a deeper
analysis and thus, to provide more reliable conclusions, espe-
cially for assessing the time needed to brainstorm the HAZOP
sessions. Table 2 lists the key data used to develop the model.
Hereafter, as illustrated in Table 1, the total time required to
conduct a HAZOP study is defined as follows:
TH = TP + TS + TW (1)
Finally, as discussed in Part I, modeling not only entails
simple regressions using least-square models, but involves
the following: (1) storing the regression statistics; (2) exam-
226 Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 224–233

Table 2 – Assembled data for HAZOP time estimation.

HAZOP P&IDs ME TP (h) TS (h) TW (h) TH (h)

A 14 20 16 40 6 62
B 15 21 18 48 5 71
C 22 19 20 54 8 82
D 24 23 25 60 9 94
E 25 30 30 65 13 108

Tp (h) 6
35

Time to Brainstorm Nodes (h)

Regression
95% CI 5
30 95% PI

25 4

20
3

15 S 1,37967
R-Sq 95,6%
R-Sq(adj) 94,1% 2
10
A B C D E
16 18 20 22 24 26 28
HAZOP Study
Tp (h) - (Model)
Fig. 2 – Individual value of the time to brainstorm nodes
Fig. 1 – Fitted line plot – Actual TP versus TP model.
versus HAZOP.

ining the residual diagnostics; and (3) generating prediction –

6
(PI) and confidence – (CI) intervals. Additionally, we note that
we conceived the HAZOP time-estimation model as the sim-
Time to Brainstorm Nodes (h)

plest possible, searching for the minimum imperative set of 5

predictors for best fitting the data.

4
2.1. Preparation time estimation (TP )

Fig. 1 shows the fitted line-plot of the actual Tp versus the 3

modeled Tp . The resulting equation to determine the time to
define, prepare, and organize the study is
2

TP (h) = 0.70P&IDs + 0.36ME (2) A B C D E

HAZOP Study

Table 3 is a list of the actions completed during the HAZOP Fig. 3 – Box plot of the time to brainstorm nodes versus
preparation phase. HAZOP.

2.2. Sessions Time Estimation (TW )
The session’s time estimation clearly depends on the number
of selected nodes. Although this also depends on the complex-
ity of the process (i.e., P&IDs and ME), we prefer to express
the TS factor as a function of the number of nodes (our intu-
itive point of view). We used this approach in Part I, employing
the indentified basic predictors. Fig. 2 shows the individual
time to examine nodes for each of the HAZOP studies con-
ducted. One node per HAZOP clearly presents a significant
higher value than the rest. A boxplot diagram highlights this
aspect (Fig. 3). The five points that do not follow the pattern are
the first nodes analyzed in each HAZOP. Due to their special
treatment and the extra tasks entailed during their analysis
(see Table 4), the time to examine these nodes is practically
twice that compared with subsequent nodes. Therefore, only
analyzing the time required to brainstorm the first node, the

Table 3 – List of actions included in the preparation time Table 4 – List of actions included during the analysis of
estimation (TP ). the first node.

HAZOP stage Actions Introducing the team members

Recording attendance
Definition Defining the study. Purpose, scope, and objectives Defining particulars of the scheduling: Break times, daily
Selecting the team roles and responsibilities timetable, lunch time
Explaining the HAZOP methodology to be used during the sessions
Preparation Obtaining information
Clarifying features of the software that will be used
Converting the information to a suitable form
Reviewing the purpose, scope, and objectives of the study
Defining the Risk Ranking criteria
Considering the process is to be analyzed
Organization Dividing the process into nodes Checking the documentation
Undertaking HAZOP time estimation Marking and confirming the node subdivision
Planning the study for arranging the meetings Reviewing the first node selected
 Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 224–233 227

130
3,0
Mean of Time to brainstorm Glob. Nd (h)

Regression
120 95% CI
2,5 95% PI

Actual Total HAZOP Time (h)

110
2,0
100

1,5 90

1,0 80

70
0,5 S 2,99404
60 R-Sq 98,0%
0,0 R-Sq(adj) 97,3%
A B C D E 50
HAZOP Study 70 80 90 100 110
Total HAZOP Time (h) - Model
Fig. 4 – Mean time to brainstorm the global node versus
HAZOP. Fig. 5 – Fitted line plot – actual TH versus TH model.

following results were obtained: (1) the mean value was 5.5 h;
and (2) after conducting a t-test, it is possible to ensure (with a
Finally, accounting for the assumptions made for the first
95% level of confidence) that the time required to brainstorm
nodes, and for the process and global node, the TS can be
them will take between 5.0 and 6.0 h. Thus, a conservative rule
modeled as follows:
of thumb is to consider that it will take 6 h to review the first
node.
Furthermore, although the analysis of the data does not TS (h) = 3(1 + Nd) (3)
highlight significant differences between process nodes (first
nodes already excluded) and the global node, it is preferable
It is important to stress that while the experience of the
to treat them independently. The selection procedure and the
team members directly influences on the time required to
critical examination involved with the global node are very dif-
examine the nodes, this factor has not been taken into consid-
ferent. Focusing on the features of the global node (Fig. 4), the
eration in the proposed model due to the results obtained from
following are its most important aspects: (1) the mean value
the five HAZOPs analyses. Data obtained from the five HAZOPs
of the time required to brainstorm a global node is 2.66 h and
carried out by five different team members converge, always
(2) applying a t-test, we can ensure (with a 95% level of confi-
explaining the proposed approach during the first session and
dence) that the time required to brainstorm a global node will
strictly following it; i.e., the structured proposal reduce the
take between 2.2 and 3.1 h. Again, the needed time increases
inherent subjectivity present in the human behavior, thus, the
with total amount of equipment that defines the process.
team members influence.
Excluding both first node and the global one from this data
analysis, the following statement applies to the process nodes
of the five HAZOPs analyzed. First, we focus on the individual 2.3. Writing Time Estimation (TW )
value of the time required to brainstorm the process nodes
versus the HAZOP study; Table 5 shows the findings from We modeled the time for documenting the HAZOP study (TW )
t-tests giving the mean time and the mean-95%-confidence as a function of TP according to previous work by Freeman
interval for these five HAZOPs. Considering the closeness of et al. (1992) and Khan and Abbasi (1997). Eq. (4) denotes the
the interval for all HAZOPs, a realistic assessment is that time required for revising the information generated during
reviewing the process nodes of each (and the global node) will the sessions up to handing out the first HAZOP draft:
occupy 3 h.

TW (h) = 0.40TP (4)

Table 5 – Mean and 95% CI for the time to brainstorm

process nodes. 2.4. Defining the model: time to conduct HAZOP
HAZOP study Variable Value studies
HAZOP A Mean 2.72
95% CI (2.50–2.93)
The final expression, in Eq. (5), represents the total expected
time to conduct a HAZOP study. It was obtained by reorga-
HAZOP B Mean 2.88
nizing Eqs. (2)–(4) according to the Eq. (1) (and taking into
95% CI (2.57–3.20)
consideration the expression obtained in Part I of the present
HAZOP C Mean 2.99 paper, i.e., the predicted nodes as a function of P&IDs and ME),
95% CI (2.78–3.20)
and after rounding off the obtained expression. The rounding-
HAZOP D Mean 3.10 off criterion was accounted for by considering that the final
95% CI (2.90–3.30) expression fits inside the 95% confidence interval (Fig. 5). This
HAZOP E Mean 3.24 procedure yielded the following easy-to-remember and useful
95% CI (3.02–3.46) equation for team leaders:
All HAZOPs Mean 3.01
95% CI (2.91–3.11)
TH (h) = 3 + 2(P&IDs + ME) (5)
228 Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 224–233
Table 6 – Planning the study for arranging meetings. Criteria for transferring hours into weeks.
Time (hours) Time (weeks)
TP (h) TP (w) = 1/16 TP (h)
TP (h) TS (w) = 1/24 TS (h) if Nd > 9
TS (w) = 1/30 TS (h) if Nd ≤ 9
TW (h) TW (w) = 1/16 TW (h)
TH (h) TH = TP + TS + TW
3. Planning the study for arranging
meetings
Having estimated the total hours for reviewing nodes, the
number of sessions is determined by dividing the total criti-
cal examination hours by hours per session. HAZOP schedules
should not be viewed as rigid timetables, but, at the beginning
of the study, they assist team members in planning their other
professional duties. They also ensure management is aware of
the required commitment for the study. The authors’ experi-
ence highlighted some guidance on how to distribute the total
hours required for conducting HAZOPs (Table 6). Thus, for the
HAZOP stages of managing and writing, and bearing in mind
that the team are highly motivated and flexible, the authors
considered the following schedule: (1) four days per week and
(2) half of the work hours per day (since members have other
daily professional duties, the time devoted to HAZOPs was half
of the work hours). However, for the examination of sessions,
six-hour meetings are considered best. Allowing this time will
ensure the review of at least two nodes per session, or only one
if it is the first node. Thus, two courses of action are appropri-
ate: (1) if nine or fewer nodes were selected, the team leader
will arrange consecutive meetings, viz., five per week ensures
that the study of nine nodes is completed within one week
and (2) if ten or more nodes were chosen, the team leader
will arrange four sessions per week, leaving time for HAZOP
members’ other work.
The criteria for planning the study clearly prioritize the
HAZOP requirements, but also afford flexibility. Furthermore,
if the sessions are consecutive, team members become famil-
iar within the methodology and the critical examination
procedure. Re-starting the HAZOP several days after the last
node reviewed, will involve giving further guidance to the pro-
fessionals involved.
4. The Deviations Structural Hierarchy
Table 1 shows that the HAZOP time-estimation model depends
on how the sessions were brainstormed. We analyzed these
sessions following the criteria established in the proposed
approach termed Deviations Structural Hierarchy (DSH). This
methodology entails two aspects: (1) considering the mini-
mum set of deviations to be brainstormed and (2) bearing in
mind their order of application. The DSH ensures the following
two aspects. Firstly, it reduces subjectivity by systematically
applying the minimum set of deviations that not only ensure
the pursued hazard identification level, but also comply with
related regulations. Secondly, it increases the level of inherent
HAZOP-structured features by defining the order of execu-
tion of the selected deviations. We note that whereas human
judgment in brainstorming generally is considered a favor-
able aspect, a well-matched methodology will guide HAZOP
members more effectively in executing a HAZOP. Finally, it is
important to stress that team leaders and expert teams will
Considerations
4 h per day; 4 days per week
6 h per session; 4 sessions per week
6 h per session; 5 sessions per week
4 h per day; 4 days per week
Sum of the total time predicted
reconvene further deviations according to a specific HAZOP’s
scope, purposes and objectives, and process nature.
4.1. Generating the appropriate level of detail for
identifying hazards
The minimum set of deviations considered should cover and
stimulate most of possible departures from the typical design
intents of the nodes. Although it is not possible to ensure
that all the hazardous scenarios will be identified, there is a
reasonable degree of consistency and uniformity in the char-
acteristics of continuous chemical processes, a fact that favors
DSH development. Moreover, the DSH keeps the global node
concept in view, as we illustrated in Part I of this paper, thereby
complementing and completing the identification of hazards
in the process nodes. Therefore, the set of deviations for ana-
lyzing process nodes include all those specific parameters that
normally are highlighted on PFDs, and also define the process-
control systems on P&IDs, e.g., – “pressure”, “level”, “phase”,
“temperature”, “flow”, and “composition”. In addition, they
also encompass the key general parameters that directly affect
the design intent, namely “human factors”, “operation”, “utili-
ties” and “containment”. A set of sixteen deviations will cover
hazard identification throughout the process nodes. In con-
trast, the deviations considered in analyzing the global node
include general parameters that will stimulate team members
to identify scenarios that could involve several nodes simul-
taneously. A set of six deviations will assure a thorough and
complete hazard identification throughout the global node.
Tables 7 and 8 illustrate the DSH methodology.
4.2. Justifying the order of reviewing deviations
Reviewing deviations requires a comprehensive definition of
the nodes. In this sense, the nodes will be numbered and
defined, clearly and unequivocally, so avoiding erroneously
interpreting their specific extension. Additionally, the inten-
tion of the parameter will be defined as completely as possible,
so that they can be classified into two categories: (1) specific
parameters, viz., variables describing physical – and chemi-
cal aspects of the process that can be measured, or detected.
The outcome of deviations on predetermined design condi-
tions could cause a hazardous scenario in the absence of
safeguards (such as flow, temperature controls) and (2) general
parameters, viz., a group of situations that, by their absence or
qualitative modifications may engender a hazardous process
condition. Although they are the causes of specific param-
eter deviations, their independent analysis is essential for
ensuring the desired level of detail in hazard identification
(additions, containment, and the like).
Focusing on continuous chemical processes, in tedious
and laborious sessions, the team should start by delineating
departures from specific parameters for two sound reasons:
Their prior identification and the analysis of relevant initiating
events (causes) gradually will assure that the team acquires a
 Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 224–233 229

Table 7 – Deviations Structural Hierarchy – Process Nodes.

Parameter Guideword Deviation Meaning

Level More More Level High level with possible overflow. Flooding of pipes
inappropriately designed for entry of liquid phase fluid
Less Less Level Low level with possible cavitations’ phenomena of pump
systems connected after vessels. Additionally, the vessel could
be totally emptied and remain dry, producing problems with
equipment designed for drying conditions

Flow No No Flow No flow in locations where flow circulation is expected

Less Less Flow Restriction of the fluid flow. Often it is analyzed jointly with
the “No Flow” deviation
More More Flow Excessive flow in fluid transport piping systems than expected
Reverse Reverse Flow The fluid circulates in the wrong direction

Pressure More More Pressure High pressure with possible hazards to the mechanical
integrity of the node equipment if design pressure is exceeded
Less Less Pressure Hazardous scenarios due to cavitation in piping, blockage in
vessels, etc. Risk of implosions in vessels of equipment not
designed for these conditions

Temperature More More Temperature High temperature with possible risk thermal damage risk in
construction materials, or possibility of activating undesirable
reactions
Less Less Temperature Low temperature with possible risk of fluid freezing,
construction materials becoming brittle or undesirable
crystallizations

Composition Other than Other than Composition Presence of undesirable chemicals due to contamination or
wrong addition (another product or impurities), by services
access (water, oil, steam) and undesirable atmospheres
generation (air)

Phase Other than Other than Phase Presence of an unintended phase state (solid, liquid or gaseous)

Utilities No No Utilities Unexpected services failure. Experience suggests considering

services analysis as a specific deviation identifies hazardous
scenarios not apparent in previous deviations

Operation As well as As well as Operation Analysis of other modes of operation than normal (start-up,
shutdown, etc.). Valuable in treating Procedural HAZOPs for
operations carried out under written instructions

Containment No No Containment Identification of any operative condition able to cause the

emission of hazardous materials off-site, such as opening
valves connected directly to the atmosphere, leaks through
joints, or breakable mechanical elements

Human factor As well as As well as Human Factor Determination whether all significant human failures are
identified by considering all the people who are involved, the
various functions they perform, and the different types of
mistakes they may make

deep knowledge of the nodes (e.g., process equipment inter-
actions, adjacent nodes relationships). This fact will assist
them in identifying future, but less obvious hazardous sce-
narios caused by the presence of complex process interactions
(e.g., “human factors”). This is the same reason why we ana-
lyzed the global node after the process nodes. Therefore, the
DSH is clearly classified and explained by three levels of detail:
(1) consider parameters’ category first; (2) establish priority of
parameters within each category; and (3) arrange deviations
within each parameter by combining with the appropriate
guideword.
4.2.1. Deviations from specific parameters
HAZOP methodology should follow the entire route of the
process fluids, mostly liquids, and gases or vapors, and a
few solids, throughout the process being reviewed. Continu-
ous chemical processes are built with a superabundance of
pipelines, pieces of equipment and instrumentation (P&IDs)
that generate the appropriate conditions for achieving new
products with more added value than simple feeds. The phys-
ical and/or chemical conditions are intentionally adjusted
or altered for meeting the processes’ sub-aims, the design
intents, and thus favor the possibility of focusing on key spe-
cific process parameters when reviewing deviations. Thus, the
first prioritization must take into account relevant parame-
ters according to process engineering; deviations generated
from specific categories of parameters initially are applied, a
fact that helps in two ways: (1) it allows members to acquire
meticulous knowledge throughout process-engineering fea-
tures and (2) it identifies key “causes” at the beginning of the
nodes analysis.
Specific parameters share inherent features, an aspect that
is highlighted by comparing their deviations (e.g., “more flow”
could lead to “more level”, and “more temperature” could
lead to “more pressure”). Therefore, three pairs of param-
eters should be considered for sequential reviewing since
they will reconcile the critical examination and hence, uni-
formity: (1) “level-flow”, (2) “pressure-temperature”, and, (3)
230 Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 224–233

Table 8 – Deviations Structural Hierarchy – Global Node.

Parameter Guideword Deviation Meaning

Utilities No No utilities Utilities failure considering the whole unit. Experience suggests
this analysis identifies hazardous scenarios not identified in
previous deviations
Operation As well as As well as operation Intended to identify modes of operation that might cause new
hazardous scenarios overlooked in previous parameters. Valuable
in treating Procedural HAZOPs for operations carried out under
written instructions
Containment NO No containment Identification of any operative condition that might cause the
emission of hazardous materials off-site focusing the analysis as a
“bird’s eye view”. Identification of direct ignition sources and
specific locations where toxics leaks are possible
Implantation As well as As well as implementation Analysis of aspects affecting facility sitng: Accessibility, get away
routes, congestion equipment, vehicle impacts, drainage systems,
slope, fire-protection devices, facility location, process units
spacing, spacing between equipment and potential ignition
sources, domino effects, emergency response issues, hazardous
area classifications
External event As well as As well As external event Analysis of aspects apart from process hazards that might
hazardous scenarios for operational safety. Consider as minimum:
lightning, flooding, deep cold, external fire and domino effect
Human factor As well as As well as human factor Analysis of aspects focused on health and safety deficiencies of
plant personnel as well as rthe occurrence of major accidents
caused for human factors n critical operations. Consider problems
limiting personnel’s activities, and the emergency procedures

“composition-phase”. The main aim here is to avoid changing
the focus of interest during the examination. The last pro-
cess rests on the following criteria: (1) first priority is following
fluid routes according to the concept of continuous chem-
ical processes (“level-flow”); additionally, emphasize “level”
rather than “flow” because nearly all of the “causes” of their
deviations are similar or comparable when reviewing “flow”
departures; (2) the second priority is following changes in the
properties of key fluids (“pressure-temperature”), and addi-
tionally, emphasizing “pressure” rather than “temperature”,
so connecting “level-flow” and “pressure-temperature”. The
findings will be more consistent by treating “pressure” after
“flow” rather than “temperature”, especially if gases or vapors
are involved; (3) the last priority is for the “composition” and
“phase” parameters, their treatment will be assisted partially
by the previous results. These priority criteria also fit with
the key P&IDs features of continuous chemical processes, and
the diagrams plenty of Level Control Valve (LCV), Flow Control
Valves (FCV), Pressure Control Valves (PCV), and Temperature
Control Valves (TCV). There also are other controls, such as
composition analyzers, that have a smaller effect. For the
“composition” and “phase” parameters only a single guide
word, “other than” must be considered in generating a unique
deviation; however, the rest of specific parameters have more
than one. Therefore, it is necessary to order them into a defi-
nite hierarchy. The DSH was established for reviewing quickly
those deviations that might entail forceful departures from
design intents. Measurable aspects, as follows, involve the
number of control systems focused on resolving these issues.
Thus (1) “level”; more hazardous scenarios could arise from
“more level” than “less level” due to possible overflow, (2)
“flow”; in continuous chemical processes fluids must move
continuously, so then the strongest departure from design
intent will be “no flow”, followed by “less flow” that sometimes
is included in the “no flow” deviation. Hereafter, the oppo-
site situation is considered, that is “more flow” that raised the
possibility of an overflow, and finally, “reverse flow”; (3) “pres-
sure” and “temperature”; the same criteria are applicable to
both parameters and they correspond to those used for “level”
assumptions by first applying “more”. Again, worse scenarios

Table 9 – Additional deviations that could complement the Deviations Structural Hierarchy.
More Less None Reverse Part of As well as Other than

Flow High flow Low flow No flow Back flow – – Loss of

containment
Pressure High pressure Low pressure Vacuum – – HP/LP interface –
Temperature High Low temperature Cryogenic – – – –
temperature
Level High level Low level No level – – Liq/Liq interface Loss of
containment
Composition Additional Loss of phase – Change of state Wrong Contaminants Wrong
or state phase concentration material
Reaction High RXN Low RXN rate No Reaction Reverse reaction Incomplete Side Reaction Wrong
(RXN) rate reaction reaction
Time Too long Too short Not started Side reaction Wrong time
Sequence Step too late Step too early Step left out Step backwards Part of step Extra action Included Wrong action
left out taken
 Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 224–233
could arise from “more” than from “less” such as, runaway
reactions, and failures of mechanical integrity.
4.2.2. Deviations from general parameters
Since each general parameter only generates a single devia-
tion, both the general parameters and the deviations hierarchy
coincide. The criteria of the order of execution were estab-
lished. Priority is given to those parameters that are more
closely connected to process engineering equipment, the
P&IDs, such as “utilities” and “operation”, so linking the
critical examination of specific parameters and the general
ones. We note that the review of “containment” and “human
factors” are more independent than those of the factors pre-
viously discussed; “human factors” is reviewed last to take
advantage of all the acquired information on nodes. Finally,
we highlight the fact that the general parameters were con-
sidered twice. Firstly, they were explored in focusing on what
could happen “inside the line” when reviewing process nodes,
and second, in identifying hazardous scenarios from a “birds
eye view” when reviewing the global node.
4.3. Deviations assessment
According to Part I of our paper, the designation of process
nodes should involve a set of equipment sharing design intent.
Therefore, it is important to refer to which point or equip-
ment is being analyzed when a deviation is being revised, for
example, heat exchangers wherein the tube and shell sides
should be assessed separately. Although several nodes will not
involve more than one point of reference, nor will it be neces-
sary to break deviations in more detailed points, it is important
to introduce this concept for ensuring the level of detail for
identifying hazards. Identified points of reference within the
actual deviation are reviewed independently, and this proce-
dure repeated as many times as there are selected points of
reference.
The meaning of deviations generated from specific param-
eters is not difficult to understand according to PFDs and
P&IDs, but this may not be the case when considering
deviations from general parameters because most are not
highlighted in the process engineering diagrams. Hence, our
discussion here is intended to offer information for under-
standing and applying general-parameter deviations.
(1) Utilities. Utilities that serve a process are considered key
parts of it because any deviations in them might occasion
the release of highly hazardous chemicals. HAZOPs treat
failures of utilities as causes of deviations, such as “no
flow”, e.g., failure of electric power to a pump. Therefore,
a thorough examination might attribute several accidents
as the consequence of a utility failure. Utilities are justi-
fied as candidates for a HAZOP parameter because of their
importance in continuous chemical processes to sustain,
cooling water, electric power, instruments air, steam, fuel-
oil, nitrogen, and because they affect more than one node.
HAZOP members might explore “No utilities” in their ses-
sions; team leader might list for independent review all
the utilities that could impact on a specific node, treating
each utility as a point of reference. We consider “No Utili-
ties” separately to take into account this deviation during
HAZOP performance and therefore, to stress its impor-
tance. After conducting the five mentioned HAZOPs, both
team leaders and team members confirmed its value.
231
(2) Operation. This category refers to other process states than
normal operation that should be addressed for identifying
hazards. According to the scope, purpose and objectives
of the study at least the startup and shutdown modes
should be examined. We propose a specific deviation, “as
well as operation” that is a formal attempt to consider
other modes. Again, after the team leader list all modes
of operation for review, they will be brainstormed sequen-
tially and independently within the mentioned deviation,
treating each mode as a point of reference. However, it is
important to stress that if the main intention of the HAZOP
study is to deeply review other modes of operation rather
than normal operation, the proposed deviation will not
cover the critical examination of them. It that case, we pro-
pose that an “independent” HAZOP should be conducted
to examine additional deviations applicable to the sequen-
tial steps of shut-down or start-up procedures. Hence, the
proposed “as well as operation” guides the decision if extra
effort is required for critically examining other modes of
operation.
(3) Containment. Despite the fact that specific parame-
ters should address loss-of-containment scenarios, we
introduced a specific deviation, “no containment”, for
researching those that might lead to fires, explosions and
toxic releases. These scenarios could be generated directly
(e.g., operator opens a valve), or indirectly (e.g., decomposi-
tion reaction resulting in releases through a pressure-relief
device). Moreover, these scenarios could be used directly
as input data for more detailed studies, such as Layer Of
Protection Analysis – (LOPA), or Quantitative Risk Analysis
(QRA).
(4) Lay-out. Although both the Occupational Safety and Health
Administration (OSHA) Process Safety Management Rule,
PSM (OSHA, 1992), and the Seveso Directive (2003) regula-
tions establish the requirement to address “facility siting”
during PHA development, there is no detailed information
on how to proceed. Attending to mandatory requirements,
but also strengthening hazard identification, the global
node includes a specific deviation, “as well as lay-out” for
HAZOP members to consider in reviewing the layout of
processes. Thus, members must review the facility lay-
out diagrams for assessing the proper location of safety
systems, including firewater headers, hydrants, extin-
guishers, and the monitors and alarms for flammable
materials and toxic gases. Additionally, the team leader
should guide the team in identifying the locations of peo-
ple and significant inventories of hazardous materials and
determine if their proximity creates potential problems
(Primatech, 2007).
(5) External events. Events originating outside the process that
might impact it adversely also must be addressed. “As well
as external event” highlights the importance of such ini-
tiating events, both when reviewing process nodes and
the global node. Flooding, vehicular impacts, and fires
or explosions in adjacent facilities are some examples of
external events.
(6) Human factors. This parameter generates the “as well as
human factor” deviation for examining on how human
factors interface with engineered process. This devia-
tion is reviewed after examining all previous deviations,
thereby ensuring that there is a deep knowledge about
node particularities. However, whereas there is some guid-
ance for treating other deviations, here there is none.
OSHA’s PSM and Seveso’s Directive regulations stress
232 Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 224–233
the need to address human factors, but information on
how to proceed is lacking, though a sound body of data
exists in nuclear safety reports written by Brookhaven
National Laboratory for the US Nuclear Regulatory Com-
mission (U.S. NRC, 2007, Higgins et al., 2007 and O’Hara
et al., 20001,2 ). The importance of assessing these human
aspects is compelling; between 50 and 90 percent of oper-
ational risk is attributable to human error (Baybutt, 2002).
As a final thought, considering the deviations of general
parameters, it is clear that HAZOP studies are eased if previous
hazard studies 1 and 2 have been previously completed and
results incorporated in the preparation for the HAZOP study.
The critical examination provided by the structural nature
of HAZOP enables the identification of hazardous scenarios
missed in previous hazard studies, both for scenarios coming
from specific and from general parameters.
5. Discussion and conclusions
We developed a mathematical model to predict the time
required to perform a HAZOP study, based upon the number
of P&IDs and the amount of major equipment in the process
facility. These are acquired and managed simply, they define
precisely the complexity of the process, and the results gener-
ated by the model are accurate. The following key results were
obtained from analyzing data collected from five HAZOPs by
following the model’s criteria:
(1) The mean-value to brainstorm first nodes, i.e., 5.5 h, clearly
is longer than the times required for other nodes (i.e., pro-
cess and global nodes). Applying a t-test showed we can
ensure (with a 95% confidence) that the time required to
brainstorm the first node of a HAZOP will take between 5
and 6 h.
(2) The mean value for the time required to brainstorm the
global node was 2.7 h; a t-test demonstrated that the time
needed to brainstorm a global node will be between 2.2
and 3.1 h. For a process node the mean value is 3.0, with a
95% confidence-interval of 2.9 and 3.1 h. There is no signifi-
cant statistical difference between the global- and process
nodes, and the close confidence interval shows that an
almost constant time is needed to brainstorm equivalent
node sizes.
We also developed a Deviations Structural Hierarchy (DSH),
a standardized procedure for examining nodes. Its applica-
tion brings two enhancements in HAZOPs. First, two sets of
deviations (one for reviewing process nodes and the other
for analyzing the global node) encourage team members to
brainstorm for any specific departure from the design intent.
Second, DSH establishes the order of application for deviations
to avoid repetitive analysis, yet structure hazard identifica-
tion without losing freedom for creative thinking. DSH is
conformed by 22 deviations that encourage questioning all
factors specified both in OSHA PSM and the Seveso Directive
for addressing hazard identification. Thus, the team leader
will select the first node to be reviewed, and after clarify
1
http://www.nrc.gov/reading-rm/doc-
collections/nuregs/staff/sr1764/r1/.
2
http://www.bnl.gov/humanfactors/files/pdf/NUREG CR-
6633.pdf.
its boundaries and its intention, he/she will follow the lay-
out of the deviations established in Table 7 for examining
the selected process node. Once the whole set of process
nodes have been reviewed by iteratively applying the contents
and criteria present in Table 7, the team leader will finally
select the global node, and will take into account the lay-
out of the deviations established in Table 8. Following this
procedure not only reinforces the HAZOP structure itself,
but also ensures reviewing the key factors that could entail
hazardous scenarios. Applying the DSH justifies the time-
estimation model proposed in the present paper and has been
validated by performing five HAZOPs in petroleum-refining
processes.
We note that Tables 7 and 8 should be used with caution.
The success of HAZOP is due to the creativity of teams work-
ing within the HAZOP methodology. This requires them to be
imaginative, and couple parameters with the help of guide-
words to identify possible deviations for the system they study.
In this sense, creative thinking during HAZOP is important.
We propose the DSH based on the results obtained from five
independent HAZOP studies. Additional deviations that may
be applicable in other studies are listed in Table 9; these devi-
ations should complement the DSH illustrated in the present
paper.
The criteria proposed in Parts I and II this paper have
been summarized by using tables, figures and “ready-to-
use” checklists for project managers, team leaders and
experts for managing, conducting and attending HAZOP
studies in continuous chemical process. These specific
“ready-to-use” tools are intended to be a reference book
for helping HAZOP responsible parties and can be found
in (Dunjó, 2010).
Acknowledgments
Jordi Dunjó acknowledges financial aid from Universitat
Politècnica de Catalunya, and the opportunity to participate
and analyze the five HAZOP studies via Trámites, Informes y
Proyectos, S.L.
References
Baybutt, P., 2002. Layers of protection analysis for human factors
(LOPA-HF). Process Safety Progress 21 (2),
119–129.
Dunjó, J, 2010, Ph.D. Thesis; “New trends for conducting HAZard
& Operability (HAZOP) studies in continuous chemical
processes. Available online: http://www.tesisenxarxa.net/.
Freeman, K.A., Lee, R., McNamara, R.P., 1992. Plan HAZOP studies
with an expert system. Chemical Engineering Progress 88 (8),
28–32.
Higgins, J.C, O’Hara, J.M., Lewis, P.M., Persensky, J.J., Bongarra, J.P.,
Cooper, S.E., Parry, G.W., 2007. Energy Sciences and
Technology Department, Brookhaven National Laboratory;
Upton, NY 11973-5000; and U.S. Nuclear Regulatory
Commission Washington, DC 20555-0001.
Khan, f.I., Abbasi, S.A., 1997. TOPHAZOP: a knowledge-based
software tool for conducting HAZOP in a rapid, efficient yet
inexpensive manner. Journal of Loss Prevention in the Process
Industries 10 (5–6), 333–343.
O’Hara, J.M., Higgins, J.C., Kramer, J., 2000. Advanced Information
Systems Design: Technical Basis and Human Factors Review
Guidance (BNL Technical Report NUREG/CR-6633).
Brookhaven National Laboratory, Environmental and Systems
Engineering Division, Upton, NY, 11973.
OSHA, 1992, Occupational Safety & Health Administration
(OSHA). Process Safety Management Rule, Process Safety
 Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 224–233 233

Management of Highly Hazardous Chemicals U.S. Department
of Labor, Washington, DC, http://www.osha.gov.
Primatech Training Institute, 2007. Training, Process Hazard
Analysis for Team Leaders (HAZOP), Amsterdam, Netherlands
(unpublished).
Seveso, 2003. Directive 2003/105/EC of the European Parliament
and of the Council of 16 December 2003 amending Council
Directive 96/82/EC on the control of major-accident hazards
involving hazardous materials. Official Journal of the
European Union, L 345/97 Brussels, 31.12.2003.