Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 214–223

Contents lists available at ScienceDirect

Process Safety and Environmental Protection

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

Conducting HAZOPs in continuous chemical processes:

Part I. Criteria, tools and guidelines for selecting 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 aim of the present paper is to provide new tools and criteria for conducting HAZOPs in continuous chemical pro-
cesses (e.g., petroleum-refining processes). These are mainly based on five HAZOPs of different systems, conducted
by different teams. Its scope covers the organizational step of the HAZOP study, which principally entails selecting the
nodes to be analyzed and estimating the time required to examine them. These two aspects are focused on defining
and planning the sessions that will be necessary to complete a HAZOP study. Part I of this paper focuses on develop-
ing tools, guidelines, and criteria for selecting and sizing nodes. A methodology for selecting nodes is illustrated and
mathematical models for predicting the number of nodes to be selected and the time needed to select each node are
presented. Part II focuses on developing a HAZOP time estimation model for predicting the time required to perform
the whole HAZOP, from its preparation and organization up to the release of the first HAZOP draft report.
© 2011 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

Keywords: HAZOP; Nodes Selection Methodology; Nodes Selection Modeling; Process Hazard Analysis

1. Introduction sequences of related accident scenarios using a rigorous and

intensive critical examination approach. Finally, HAZOP iden-
Process Hazard Analyses (PHAs) techniques are conducted for tifies protections that were already part of the design which
identifying hazards in new or existing process facilities. PHAs may prevent the accident, and if safeguards are insufficient to
are designed not only to ensure the safe design and opera- solve the problem, recommendations should be considered.
tion of a system, but also the complete risk evaluations and Further key HAZOP related information can be found in Lawley
adequate protection devices. In this sense, PHAs could be (1974, 1976), CIA (1977), Knowlton (1981), CCPS (1992), Nolan
considered the foundation of process safety and Risk Man- (1994), Lees (1996), Wells (1996), Kletz (1999), EPSC (2000), BS
agement Programs (RMPs). The HAZard & OPerability (HAZOP) IEC 61882 (2001), and (MacDonald, 2004). Additionally, Dunjó
study is a PHA applied worldwide, which is used as a method et al. (2010) published a paper on the state-of-the-art, the cur-
for studying not only hazards, but also operability problems rent limitations, new trends and research needs of a HAZOP
of a system, by exploring the effects of any deviations from study.
design conditions. This method reviews equipment, instru- The aim of the present paper is to provide new tools
mentation, utilities, human factors and external events that and criteria for conducting HAZOPs in continuous chemical
might impact the process by following Piping & Instrumenta- processes (e.g., petroleum-refining processes). Understanding
tion Diagrams (P&IDs). Breaking the design into manageable and hazoping these kinds of processes is not an easy task
sections with definite boundaries called “nodes”, HAZOP due to their complexity and wide range of equipment, instru-
techniques identify hazards and qualitative causes and con- mentation, utilities, etc. with multifarious interrelations to be

∗
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.2011.03.001
 Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 214–223 215

Table 1 – Key HAZOP documentation for organization

Nomenclature 1 purposes.
Symbol Documentation Description
Exchangers number of exchangers Process description Detailed written description of
FCVs number of flow control valves process principal sections
LCVs number of level control valves Process Flow Diagrams Relationship between major
ME number of major equipment (PFDs) equipment of a plant facility
MH total man hours (h) Piping & Instrumentation Relationship between process
Diagrams (P&IDs) equipment and control
N number of sessions required to conduct a
instrumentation
HAZOP
Plot plan Schematic description of the
Nd number of nodes equipment layout
Nd/PS number of nodes per principal section Previous HAZOPs Consideration of previous HAZOPs
PCVs number of pressure control valves conducted on the same process
PDescription Number of process description pages Corporative HAZOP Corporative HAZOP guidelines to
PFDs number of process flow diagrams guidelinesa follow

P&IDs number of piping & instrumentation diagrams a

Not all companies have developed its own guidelines.
PS number of principal sections
Pumps number of pumps
TCVs number of temperature control valves 2. The HAZOP organization phase
TBTNd time required to examine a node (*h)
TH total time required to conduct the HAZOP study The starting point for organizing complex HAZOP studies is to
(h or **w) be aware of the key needed input data (see Table 1), and which
TNd total time required to select nodes (***min) results are required for ensuring a well-suited organization
TNdPFD time to select Nodes on PFDs (min) (see Table 2). Thus, the topic which is necessary to introduce
TNdP&ID time to transfer Nodes from PFDs to P&IDs (min) as soon as possible is the concept of node, which will affect
TP total time required to prepare and organize the directly not only the planning of the study, but also the level
study (h) of resolution of the hazard identification.
TPS time to select principal sections on PFDs (min)
TS total time required to execute the study (h) 2.1. Dividing the design into manageable sections –
TW total time required to write the draft report (h) nodes definition

The HAZOP documentation for hazard identification is based

reviewed. In this sense, the paper scope covers the organiza- on P&IDs, but it is not possible to review the design by con-
tion step of the HAZOP study, which principally entails the sidering the information contained in these diagrams as a
node selection, and the time which will be required to exam- whole (by doing so, most of hazardous scenarios would be
ine them. These two aspects are totally focused on defining missed). Thus, the design has to be broken into manageable
and planning the sessions that will be necessary to finalize sections with definite boundaries (nodes) for detailed review.
the HAZOP study with success. Regarding the aspects to be The HAZOP methodology will be applied for analyzing inde-
covered, the paper has been divided in two parts, which both pendently each design intent section, this means how the
of them are totally focused on defining the organization step process section is expected to be operated. Then, the method-
of the HAZOP study with the aim to develop criteria for plan- ology will generate deviations (departures from design intents)
ning the HAZOP sessions and thereafter ensuring a well-suited by combining guide words (simple words or phrases used to
support for effective hazard identification. qualify or quantify design intents (e.g., more, less, other than))
On one hand, Part I of the present paper focuses on develop- and parameters (aspects that describe design intents (e.g.,
ing tools, guidelines and criteria for selecting and sizing nodes. temperature, composition, level)) for each node. The objective
Not only a methodology for selecting nodes will be illustrated is to explore possible causes, scenarios and safeguards, and
(i.e., Nodes Selection Methodology (NSM)), but also a mathe- if necessary, to decide whether action is warranted. For this
matical model design to predict the number of expected nodes
to be selected will be presented. Finally, an additional model
Table 2 – Pursued results for HAZOP organization.
able to predict the time needed to select those nodes will be
illustrated by following the mentioned NSM. Variable Units Description
On the other hand, part II of the present paper will focus Nodes features – Level of detail for hazard
on developing a HAZOP time estimation model, able to predict identification
the time needed to perform the whole HAZOP-related steps, Number of nodes – Total number of nodes to be
from its preparation and organization up to the release of the selected
first HAZOP draft report. This model will take into account HAZOP Time Hours Required time for managing,
executing and writing (1st
the required time for examining the selected nodes and it
draft) the study
will include the criteria that have been followed during their Number of sessionsa – Total number of sessions for
analysis. These criteria are based on a new approach called arranging meetings
Deviations Structural Hierarchy (DSH). Total elapsed timea Weeks Time distribution according to
study plan

a
1 Subject to criteria for planning the study.
*h: hours; **w: weeks; ***min: minutes.
216 Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 214–223

reason, a key HAZOP decision is how to break the design into pressure”, “more temperature”, etc. will be independently
sections for ensuring attuned hazard identification (i.e., choos- applied in each node by following the traditional guide-word
ing level of detail for hazard identification). Even though the approach. The global node is a single node which is based on
importance of the nodes selection, no research has been con- the Plot Plan Diagram and is intended to complement pro-
ducted for establishing standard criteria for it, and nowadays cess nodes identification abilities by focusing attention on
it is considered partly as an art. what could happen “outside the line” as a “bird’s-eye view”.
The research was focused not only on detecting well- Deviations such as “loss of containment”, “as well as human
delimited sub-design intents from the main process intention, factor”, etc. will be applied. The author’s experience confirms
but also on deciding the optimum sizes for ensuring that the the effectiveness of using the global node on detecting haz-
HAZOP methodology would be satisfactorily applied. There- ardous scenarios that could be omitted by only using process
fore, all section elements have to share the same design intent, nodes; Dunjó (2010): (1) initiating events that affect more than
and all sections have to be equivalent-sized. In one hand, if one node (e.g., flooding, utilities), (2) specific issues that affect
large nodes are selected, the expert team will become mired, more than one node (e.g., facility siting, human factors, pip-
and will make hazard identification difficult and the analysis ing issues), (3) need to look at hazards from the perspective
more confusing. On the other hand, if very small nodes are of the overall process (e.g., multiple failure scenarios may
selected, the study will become repetitive, the general picture involve causes originated from more than one node), and (4)
of the process will be harder to understand, and the study will the necessity to take into account the well-known domino
become extremely time-consuming as well. effect concept.
HAZOP members should be aware of two representations We note that the questions related to the global node
with complementary scopes: (1) logical representations which should be reviewed in earlier PHA rather than HAZOP.
show relationships between parts of the process without scale However, we consider this to be a complementary anal-
(P&IDs) and (2) physical representations which are intended ysis, and always after “hazoping” the process via the
to represent the actual process layout (Plot Plans). Hereafter, “process nodes”, it will be helpful to think again on vari-
nodes were also classified into two categories: the traditional ables such as domino effect, facility sitting, and other
process nodes and the global node, a new approach which has broad effects. Considering such consequences, is also help-
been welcomed by HAZOP members. Furthermore, only con- ful for Quantitative Risk Analyses (QRAs), and Consequence
sidering both kind of nodes it is possible to ensure the review modeling.
of the all factors addressed in the “Seveso” Directive (2003) and
in the OSHA PSM Rule (1992) when taking into account the Pro- 3. Nodes Selection Methodology – NSM
cess Hazard Analysis (PHA) phase of any Safety Management
System (SMS).
The endeavors on developing the Nodes Selection Methodol-
ogy (NSM) were focused on engendering the most excellent
2.2. Process nodes and the global node concept chance for acquiring the desired process knowledge for lead-
ing HAZOP studies.3 In these sense, it was never planned to
The emergence of new technologies and the chemical automate (via expert systems development) the nodes selec-
processes evolution introduced new routes to failure and tion. According to the authors’ judgment, human decisions
potential risks. Therefore, not only it is important to decide could introduce bias, but undoubtedly, they are the key aspect
which new deviations should be applied for analysis, but to enrich hazard identification studies. The main aim of the
also to be aware of both logical and physical process rep- methodology is to find a set of equipment pieces and lines (sec-
resentations. The traditional methodology reviews logical tions/nodes) that together perform a certain process sub-task,
representations (P&IDs). However, it is also important to sharing design intent.4 These sections, if well-defined, will
review physical representations for detecting additional sce- speed up the critical examination approach avoiding repetitive
narios by applying complementary HAZOP deviations. With discussions Moreover, if well-suited deviations are applied,
the aim to ensure the most possible achievable hazard iden- the hazard identification level will be maintained.
tification level, two nodes categories2 are also contemplated: Continuous chemical processes are complex and multifar-
(1) Process nodes and (2) Global node. These categories will ious. In addition, they involve a large number of equipment,
allow reviewing process properties with two different levels instrumentation, utilities, etc, that are related through mul-
of detail and points of view. Firstly, exhaustive review via pro- tifaceted issues. Therefore, it is not easy to understand their
cess nodes, and lastly, conducting a widespread review via the specific features (e.g., design intentions). Thus, facing up to
global node. The global node will be checked at the end of the the information contained into P&IDs (sometimes tedious and
whole hazard identification procedure since if it is analyzed in difficult to understand) as a first attempt should be avoided.
first place, the expert team may be disinclined to study indi- For this reason, the main goal of Nodes Selection Methodol-
vidual nodes with the required attention. This could imply ogy is to sequentially describe which steps have to be followed
overlooking important causes that can produce hazardous and which documentation has to be analyzed. Its purpose is
scenarios; Dunjó (2010). to assist and direct team leaders going thoroughly into the
Process nodes are process sections which share design process according to a certain course of actions for acquiring
intent and are constituted by a set of lines and equipment progressively the adequate process knowledge. This knowl-
– together with their instrumentation, utilities, etc. They are
based on P&IDs and are focused on identifying what could
happen “inside the line”. Deviations such as “no flow”, “less 3
The methodology not only assess experienced leaders, but
also gives an opportunity to novice one’s for leading complex and
tedious HAZOPs.
2 4
Part II defines the minimum set of deviations and their order Design intent may refer not only to plant equipment but also
of application for both nodes categories. may cover what is needed to be done within the section.
 Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 214–223 217

edge will not only provide strength on nodes selection, but by the process description documentation. Location of con-
later on it will also advance sessions. ditions changes might be important points to consider as
The NSM starts from the main process design intention node boundaries. In this sense, control valves, pumps, etc. are
(e.g., Ethyl Tert Butyl Ether (ETBE) production) and sequen- important items to contemplate.5 To sum up, analyzing PFDs
tially guides team leaders breaking the process up to nodes. and making use of the process description, team leaders will
The minimum documentation required is Process Description, obtain two results: (1) nodes highlighted on PFDs and (2) pro-
PFDs and P&IDs. Then, the methodology will guide team lead- cess knowledge intended to help team leaders on selecting
ers on how to review that documentation for carrying out the these preliminary nodes and further decisions. Fig. 3 illus-
following steps: (1) selecting principal sections, (2) selecting trates an example of preliminary nodes marked on PFDs.
preliminary nodes on PFDs, and (3) transferring nodes from
PFDs to P&IDs. Fig. 1 illustrates the nodes selection flux dia-
gram. 3.3. NSM third step: transferring nodes from PFDs to
P&IDs
3.1. NSM first step: selecting principal sections
It is imperative to transfer nodes selected from PFDs to P&IDs.
There are several intents which are connected to create the Whereas PFDs show only major equipment relationships,
overall process. The first attempt for nodes selection is to iden- P&IDs detail all process equipment and its relationship with
tify principal sections (PSs) (unit operations) which are easy to control instrumentation, utilities, etc. Consequently, addi-
“disconnect” from each other (e.g., feeds section, reaction sec- tional equipment and process control will be involved on
tion, absorber section, distillation, etc.). Principal sections are selecting detailed nodes and sizing may change from pre-
addressed as a wide number of equipment which is involved in liminary nodes. Hereafter, team leaders will independently
achieving a sub-aim: contributing to the overall design inten- focus attention on each selected PS, and after transferring all
tion of the process. This is the first chased size for breaking preliminary nodes involved, they will iteratively repeat the
the process into nodes. same procedure per each section. Henceforth, team leaders
Process description and PFDs have to be analyzed for will thoroughly detect the boundary equipment of a given pre-
defining and selecting PS. In one hand, process description liminary node, and will look for them on P&IDs. In some cases,
normally outlines PS which are clearly defined by detail- additional equipment will come into view, and the team leader
ing equipment, control loops, chemicals, process conditions, should evaluate if it has to be included inside the present
utilities and section boundaries. With meticulous process node, if it has to be incorporated in a contiguous node, or if
description reading, team leaders not only will be able to select it has to be considered as a new node. On one hand, if addi-
PS, but also will understand which different design intents tional set of equipment has sufficient entity to be “noded”
are present inside each of them. On the other hand, PFDs will because its size or meaningful potential hazardous scenar-
graphically provide: (1) how PS are connected and (2) which ios, team leaders will consider it as a new node. On the other
major equipment concerns each PS. Thus, after identifying the hand, if additional equipment shares the same selected node
first PS, the team leader will mark its boundaries on PFDs and design intent, equipment will be added. In case of doubt,
will iteratively repeat the same steps up to the last section con- the process description and the process knowledge acquired
sidered. To put it briefly, reading the process description and will help team leaders on taking the right decision. In con-
making use of PFD, the team leaders will obtain two results: (1) clusion, considering both PFDs and P&IDs and making use
PS highlighted on PFDs (see Fig. 2), and (2) process knowledge. of process description, team leaders will obtain two results:
(1) detailed nodes highlighted on P&IDs and (2) satisfactory
3.2. NSM second step: selecting preliminary nodes on process knowledge. The latter is very valuable information to
PFDs proceed with the sessions.

Whereas the aim of selecting PS is to break the process into

sub-aims (big-sized sections), in this step, team leaders will 4. Justifying the NSM – field work and data
decide a well-suited nodes sizing according to design intents. analysis
PS will be broken into smaller sections which will define
the final nodes sizing. The input data required for selecting After conducting several HAZOPs intended to “fine-tune” and
preliminary nodes are the process description and the PFDs define the criteria highlighted in the present paper, five well-
(diagrams which contain process information easier to under- suited HAZOPs were organized and executed following the
stand than P&IDs). criteria established in the present paper. A statistical analy-
Team leaders will independently focus attention on each sis of different variables (see Nomenclature) recorded during
selected PS, and after detecting several design intents, they these five mentioned studies helped and provided guidance
will select and mark the preliminary nodes on PFDs. This on taking decisions and conclusions on how to proceed in
approach will be iteratively repeated per each section. Accord- all HAZOP-related phases. The numerical analysis was basi-
ing to the process description, PFDs are useful to understand cally intended to assess and compare sample distributions for
visually the entire route of the fluid system through the each variable or groups of variables collected, to examine their
process. This allows the possibility to be aware of process shape and spread, and also, if possible, to find a relationship
conditions changes, key points for starting and ending nodes. between variables.
Consequently, team leaders will decide which major equip-
ment, lines, pumps and control valves will be considered
within the same node. Generally two, three, and rarely four
nodes (different intentions) will be present in one PS. They 5
Boundary items concern contiguous nodes, important fact for
should have been identified by and the team leaders helped examination interests.
218 Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 214–223

Fig. 1 – Nodes selection flux diagram.

4.1. Estimating the number of nodes to be selected present in the process (ME), which are clearly illustrated on
PFDs, and (2) number of P&IDs required to define the process
The expected number of nodes to be selected is directly influ- (P&IDs). Therefore, a simple regression using least squares was
enced by the complexity of the process. For this reason, for performed. Apart from fitting general least squares models;
modeling purposes, several variables were combined between also the following tasks were conducted: (1) storage of the
them with the aim to explain the complexity of the process regression statistics, (2) examination of the residual diagnos-
under review (i.e., number of pieces of major equipment,6 tics, and (3) generation of prediction (PI) and confidence (CI)
number of P&IDs, number of PFDs, total amount of “minor” intervals. The results of the regression are shown in Fig. 4,
equipment (e.g., FCVs, pumps) present in the process). After and the resulting equation to determine the number of nodes
many combinations (all of them studied both from the point of to be selected (Nd) is the following:
view of mathematics and process safety matters), it was pos-
sible to establish a well-suited regression between the total
number of expected nodes to be selected with the following Nd = 0.29P&IDs + 0.50ME (1)
variables (predictors): (1) number of pieces of major equipment
We note that quantifying the combination of the P&IDs and
ME variables is the simplest way to implicitly define the level
6
Major equipment refers to equipment that presents of P&IDs congestion. If a large number of P&IDs and a limited
containment, i.e., volume. Major equipment is easy to be number of ME define the process, we are working with low con-
identified via PFDs as clear unit entities normally called “unit
gested P&IDs. On the other hand, if a few P&IDs are present,
operations”, e.g., reactor, stripper, separator. It is better to
identify major equipment in PFDs rather than P&IDs due to in the but a huge number of pieces of ME define the process, P&IDs
last; additional equipment is highlighted and could lead to errors. are congested.
Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 214–223 219

Fig. 2 – Principal sections example.

Fig. 3 – Preliminary nodes example.

220 Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 214–223

95% CI for the Mean

22 Regression 30

Time to Select PS on PFDs (min)

95% CI
Number of Nodes

20 95% PI
25

18 20

16 15

14 S 0,668249 10
R-Sq 92,2%
R-Sq(adj) 89,6% 5
12
14 15 16 17 18 19 20 21 22 23 0
Number of Nodes - (Model)
A B C D E
Fig. 4 – Fitted line plot – actual Nd versus Nd model. HAZOP Study

Fig. 6 – Individual value of the time to select PS on PFDs

12 Equipment
MEs
versus HAZOP.
FCVs
10 LCVs pieces of equipment and with more than 12 should be
PCVs
TCVs revised).
8 Pumps
Exchangers
Data

6
4.2. Estimating the time for nodes selection

4 It was also possible to predict the time required to select the

expected nodes via analyzing the three steps that the NSM
2 establishes: (1) principal sections definition, (2) Preliminary
nodes marked on PFDs, and (3) Detailed nodes marked on
0 P&IDs.
HAZOP A HAZOP B HAZOP C HAZOP D HAZOP E

Fig. 5 – Mean value of equipment present per Nd and per 4.2.1. Analyzing the principal sections
HAZOP. The following section reviews data collected from the time
required to decide and to select the PS on PFDs. The key
4.1.1. Analyzing the selected nodes structure documentation that supports the PS selection is: the process
The analysis of the structure of the selected nodes aims at description contents (which can be more or less detailed), and
detailing the equipment present in the nodes selected follow- the PFDs (which can be more or less congested from the point
ing the proposed methodology. A t-test has been conducted for of view of the number of pieces of equipment present in a
each variable per HAZOP, and finally, the global results (consid- specific PFD). Thus, the ratio between number of pages used
ering the whole set of HAZOPs together) give an idea about the to describe the process (the level of detail), and the number of
dimension and structure of nodes selected. Fig. 5 shows the PFDs (a measurement of diagrams congestion) should influ-
mean value of equipment present per node and per HAZOP. All ence on the required time to select the PS (most evidence
HAZOPs present an equivalent mean number of pieces of total should be depicted if it involved a team leader without experi-
equipment per node, which fits between 8 and 8.5. Whereas ence). Figs. 6 and 7 show the individual time values to select PS
the specific equipment that constitutes a node will be differ- on PFDs, and the mean ratio between PDescription and PFDs,
ent depending on the process analyzed (e.g., HAZOP D, which respectively.
present a mean value of 10), the mean of the total amount of Comparing the time required to select PS and the PDe-
equipment that constitutes a node is coherent with regard to scription/PFDs ratio, it is assumable that if the ratio drops, the
the five HAZOPs analyzed, and also gives an idea about the time to select PS will also be lower. This fact is confirmed for
node sizing. HAZOPs B, C and D. However, in the case of HAZOPs A and E it
Therefore, and only as a rule of thumb, it is can be stated that the first one requires less time to select the
possible to state a margin of the equipment that
should be present in a specific node. Data is sum- 25
Mean of Rao (PDescripon, PFDs)

marized in Table 3 (e.g., a node with less than 5

20
Table 3 – Expected number of equipment in process
nodes. 15
Number of equipment Mean 95% CI Expected
Equipment
10
MEs 1.47 (1.30–1.63) 1–2
FCVs 2.45 (2.00–2.91) 2–3 5
LCVs 0.60 (0.39–0.81) 0–1
PCVs 1.25 (0.97–1.52) 1–2
TCVs 0.70 (0.38–1.02) 0–1 0
A B C D E
Pumps 0.88 (0.68–1.08) 0–1
HAZOP Study
Exchangers 1.65 (1.31–1.99) 1–2
Total equipment 5–12 Fig. 7 – Mean ratio of [PDescription/PFDs] versus HAZOP.
 Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 214–223 221

95% CI for the Mean 95% CI for the Mean

7 14
Time to Select Nodes on PFDs

Time to Select Nodes on P&IDs

6 12

10
5
8
4
6
3
4
2
2
1 0
A B C D E A B C D E
HAZOP Study
HAZOP Study
Fig. 9 – Individual value of the time to select Nd on P&IDs
Fig. 8 – Individual value of the time to select Nd on PFDs
versus HAZOP.
versus HAZOP.

PS, even though both studies present an equivalent number of

pages to describe the process (22 and 21 pages, respectively). be selected. It is present in a PS that contains two nodes, and
Thus, the only variable that can explain these differences is the number of pieces of major equipment that defines itself is
the number of PFDs. Whereas HAZOP A presents only one 3. According to the previous analysis, this node complies with
PFD, the HAZOP E takes three diagrams to describe the pro- “normal” featured nodes, and, it has been decided not to con-
cess. The ratio is less in HAZOP E rather than HAZOP A, but sider it as an outlier point. However, the second point (HAZOP
the time required to select PS is greater in HAZOP E. This is E) takes 7 min for its selection and its main features are the fol-
explained by introducing a new variable in the present analy- lowing: (1) it comes from a PS that only contains the mentioned
sis: the number of P&IDs, which are 14 for the HAZOP A; and node; (2) it takes 4 major pieces of equipment for its defini-
25 for the HAZOP E. The number of P&IDs is valuable infor- tion. These characteristics present divergence with the rest of
mation that takes into consideration the process complexity, selected nodes, and for this reason will be not considered for
and confirms that more time is required to understand the further analysis. Additionally, the reasoning established in the
process analyzed via HAZOP E, i.e., more time to select PS. present paragraph confirms coherence when considering the
There was an attempt to find a model for predicting the time optimum node and PS sizing in relation to the time needed to
required for selecting PS as a function of the process descrip- define them. Hereafter, a t-test applied to the time required to
tion (PDescription, PFDs, P&IDs), but no coherent correlation select preliminary nodes establish a 95% confidence interval
was found. Subjective factors such as: team leader experience, of 1.67 and 2.12 min, where the mean equals to 1.90. But after
clarity in process description contents, and also the availabil- excluding the outliner point, this interval closes to 1.63 and
ity to understand PFDs and P&IDs affect it directly. However, 1.95 min (less than 2 min). Finally, and taking into account the
and after considering the whole set of HAZOPs analyzed, a minimum time required to select preliminary nodes, a rule of
t-test performance states that, with a 95% of confidence, the thumb could state that a node will take 2 min to be selected,
time required to select an individual PS will take between 7.17 and additionally: (1) it comes from a PS that contains 2 or 3
and 11.76 min. A conservative rule of thumb could be to con- nodes, (2) the amount of major equipment present in the node
sider 10 min per PS, obviously, only for guidance and not as will be 1 or 2. Whereas no individual values are possible to
a strict rule. This should help inexperienced team leaders to exactly define these studied features, small intervals depicted
have an idea about how long will take the PS selection. Finally, (CI) not only help to understand the methodology, but also
and focusing on the number of PS selected, it could be possi- confirm the optimum size and time both for PS and nodes.
ble to establish guidance from it. The following reasoning can
be followed: (1) a t-test highlights a 95% confidence interval 4.2.3. Analyzing the detailed nodes
of (2.36–3.56) nodes per PS, i.e., mean = 2.96; and (2) a t-test This section analyzes data collected from the time required
highlights a 95% confidence interval of (1.30–1.63) of major to transfer the preliminary nodes from PFDs support to the
equipment per node, i.e., mean = 1.47. Therefore, taking as 3 P&IDs, thus defining the detailed nodes (see Fig. 9). A t-test
the mean number of nodes present in a PS, and 1.5 the mean focused on the time required to select detailed nodes states
number of major equipment per node, the team leader should a 95% confidence interval of (5.52–6.79), i.e., mean = 6.16 min.
size PS with a number of major equipment no more than 6, However, these values are susceptible to being dropped. Focus-
and no less than 3. As an optimum number, also validated ing on the individual value of time required to select detailed
via engineering judgment, well-sized PS will take 4 or 5 major nodes on P&IDs, it is evident that HAZOP D takes more time
equipment. than others. After analyzing the number of nodes that come
from the same PS, it is confirmed that HAZOP D presents the
4.2.2. Analyzing the preliminary nodes highest value compared to the others (Nd/PS ratio). As illus-
The following contents review data collected from the time trated before, it is confirmed that with a number of more
required to decide and select preliminary nodes on PFDs, thus, than 3 nodes per PS it directly increases the time to select
the time to select nodes from PS on PFD (see Fig. 8). detailed nodes. Therefore, and only for illustrative purposes,
Basic statistics highlight that two nodes could be consid- once extracted nodes that come from PS defined with more
ered as outlier points. If so, they could be taken off of the than 3 nodes, the new t-test analysis states the following
analysis. A preliminary study was done to decide on this mat- results: (1) 95% confidence interval: (5.09, 7.10) min, and (2)
ter. The first point (HAZOP A) is a node that takes 5 minutes to Mean: 6.09 min.
222 Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 214–223

300 global node based on the plot plan diagram for attending
Regression to issues that encompass more than one process node (e.g.,
Time to Select Nd / HAZOP (min)

95% CI
95% PI
utilities, flooding, facility siting, piping configuration) is also
200 instructive. The proposed Nodes Selection Methodology (NSM)
is intended to enhance for the HAZOP process by breaking
the process into optimal size nodes, and by guiding HAZOP
100 team leaders. The NSM breaks progressively the process up
to desired nodes sizing by: (1) considering the main process
principal sections (unit operations), (2) defining the pursued
0 nodes sizing on PFDs (preliminary nodes), and (3) transferring
S 13,4607
nodes from PFDs to P&IDs (detailed nodes). This sequential
R-Sq 98,4%
R-Sq(adj) 97,9% approach may serve to show the useful points, namely a rough
-100 guide with would be of potential value to inexperienced lead-
0 5 10 15 20 25 30 ers. Additionally, experienced leaders will already be aware of
Number of ME / HAZOP
the comparable results for their systems and will have devel-
Fig. 10 – Fitted line plot – time to select Nd versus ME. oped effective, and probably closely comparable, methods for
selecting nodes.
4.2.4. Defining the model: time to select detailed nodes on Our numerical analysis confirms the proposed NSM
P&IDs methodology. A simple model was developed for predicting
Once individual factors of the selection methodology have the number of nodes to be selected, and a second easy-to-use
been analyzed (i.e., PS, preliminary nodes, and detailed nodes), model was developed to predict the time required to select
it is appropriate to consider the total time required for node the nodes by following the proposed NSM. These tools, crite-
selection. The intention is to fit the total time with basic vari- ria and guidance will help team leaders on assessing how to
ables that are inherent from the process and easily taken into plan the study for arranging meetings.
account just before starting the organization of a HAZOP study. With reference to basic statistics applied to the related
The total time required to select detailed nodes on P&IDs is data, it is possible to take as 3 the mean number of nodes
defined as follows: present in a principal section and 1.5 the mean-number of
pieces of major equipment per node. Thus, the team leader
TNd = TPS + TNdPFD + TNdP&IDs (2) should size principal sections with a number of pieces of major
equipment to no more than 6, and no less than 3. As an
Here again, several coherent variables were born in mind optimum number, also validated via engineering judgment,
with the aim of modeling the total time required to select well-sized principal sections should take between 4 and 5
detailed nodes. Obviously, the number of nodes basically takes pieces of major equipment.
into account the number of equipment that constitutes the It has been checked, and validated, that the mean value
process, and principally the major equipment. For this rea- of equipment present per node and per HAZOP is statistically
son, it was easy to decide modeling the time required to select equivalent. This number ranges between 8 and 8.5. Only one
detailed nodes as a function of the number of major equip- of our examples (i.e., HAZOP D) shows a greater value than
ment present in the process under review. It is important 10, and this high number was criticized. Whereas the spe-
to mention that PS features of the HAZOP C do not com- cific equipment that constitutes a node could vary (depending
ply with the fact that they should be constituted by no more on the process analyzed), the mean total number of pieces
than 4 nodes. This is the only point that causes divergence of equipment that constitutes a node is the same in the five
when the total time required to select nodes is plotted versus HAZOPs analyzed, and gives an indication of node sizing. The
the number of major equipment present in the process to results obtained also confirm that the equipment selected (i.e.,
be “hazoped”. According to previous contents, it has been FCVs, LCVs, PCVs, TCVs, Pumps, and Exchangers) for analysis
decided to exclude and consider it as an outliner. This action is representative of the nature of the processes (continuous
improved the fitted line (see Fig. 10), which establishes the chemical processes (e.g., petroleum-refining processes)).
expected time to select nodes (see Eq. (3)): After conducting a t-test applied to the time required to
select preliminary nodes, results confirm a 95% confidence
TNd (min) = 8 ME (3) interval of 1.63 and 1.95 min, where the mean equals 1.87. A
rule of thumb could state that a node will take 2 min to be
We note that whereas the time required to select nodes selected on PFDs, and additionally: (1) it should come from
may not be significant relative to the time if takes to complete a principal section that contains 2 or 3 nodes, (2) the num-
a HAZOP, this time factor is useful for estimating the level of ber of pieces of major equipment present in the node will be
effort and budget of a project. 1 or 2. Whereas no individual values are possible to exactly
define these studied features, the small intervals provided
5. Conclusions (confidence intervals – CIs) not only help to understand the
methodology, but also tackle the optimum size and time both
Criteria and tools are proposed for assisting HAZOP nodes’ for principal sections and nodes.
management. We offer guidance on how to select nodes, and Additionally, in the case of nodes that come from princi-
which node size is the optimum one for HAZOP performance. pal sections defined with more than 3 nodes, a t-test analysis
Such selection is conducted by ensuring the desired level of states the following results about the time required to select
detail for identifying the potentially most hazardous scenar- nodes on P&IDs: (1) 95% confidence interval: (5.09, 7.10) min,
ios. This goal is supported by generating process knowledge and (2) Mean: 6.09 min. Thus, defining principal sections with
via analyzing key process information. The addition of the no more than 3 nodes, as well as defining nodes with no
 Process Safety and Environmental Protection 8 9 ( 2 0 1 1 ) 214–223 223

more than 2 major pieces of equipment are the two factors Dunjó, J, 2010. PhD Thesis, “New trends for conducting HAZard &
than ensure the minimum time required to select nodes (i.e., Operability (HAZOP) studies in continuous chemical
principal sections, preliminary nodes, and detailed nodes). processes. Available online: http://www.tesisenxarxa.net/.
Dunjó, J., Fthenakis, V., Vílchez, J.A., Arnaldos, J., 2010. Hazard
Finally, and only for guidance purposes, a model has been
and operability (HAZOP) analysis. A literature review. Journal
developed to predict the time that a node selection will take. of Hazardous Materials 173 (1–3), 19–32.
The response is a function of the number of pieces of major EPSC, 2000. HAZOP: Guide to Best Practice. European Process
equipment that constitutes the process to be hazoped. Safety Centre, Institution of Chemical Engineers (IChemE),
Rugby, UK.
Acknowledgments Knowlton, R.E., 1981. Hazards and Operability Studies, the
Guideword Approach. Chemetics International Company,
Vancouver.
Jordi Dunjó acknowledges financial aid from Universitat
Kletz, T.A., 1999. HAZOP & HAZAN: Identifying and Assessing
Politècnica de Catalunya, and the opportunity to participate Process Industry Hazards, Fourth ed. Institution of Chemical
and analyze the five HAZOP studies via Trámites, Informes y Engineers, Rugby, UK.
Proyectos, S.L. Lawley, H.G., 1974. Operability studies and hazard analysis.
Chemical Engineering Progress 70 (4),
45–56.
References
Lawley, H.G., 1976. Size up plant hazards this way. Hydrocarbon
Processing 55 (4), 247–261.
BS IEC 61882, 2001. Hazard and Operability Studies (HAZOP Lees, F.P., 1996. Loss Prevention in Process Industries: Hazard
Studies) – Application Guide. International Electrotechnical Identification, Assessment and Control, Second ed.
Commission. Butterworths-Heinemann, Oxford, UK.
CCPS, 1992. Guidelines for Hazard Evaluation Procedures. In: Macdonald, D., 2004. Practical HAZOPs, Trips and Alarms.
Second Edition with Worked Examples. Centre for Chemical Newnes Publications, Burlington.
Process Safety, American for Chemical Engineers, New York. Nolan, D.P., 1994. Application of HAZOP and What—If Safety
CIA, 1977. A Guide to Hazard and Operability Studies. Imperial Reviews to the Petroleum, Petrochemical and Chemical
Chemical Industries and Chemical Industries Associations Industries. Noyes Publications, New Jersey.
Ltd., London, UK. OSHA, 1992. Occupational Safety & Health Administration
Directive, 2003/105/EC of the European Parliament and of the (OSHA). Process Safety Management Rule, Process Safety
Council of 16 December 2003 amending Council Directive Management of Highly Hazardous Chemicals. U.S.
96/82/EC on the control of major-accident hazards involving Department of Labor, Washington, DC, http://www.osha.gov.
hazardous materials. Official Journal of the European Union, L Wells, G., 1996. Hazard Identification and Risk Assessment.
345/97 Brussels, 31.12.2003. Institution of Chemical Engineers (IChemE), Rugby, UK.