You are on page 1of 12

Journal of Loss Prevention in the Process Industries 26 (2013) 1015e1026

Contents lists available at SciVerse ScienceDirect

Journal of Loss Prevention in the Process Industries


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

Assessment and improvement of integrated HSE and macro-


ergonomics factors by fuzzy cognitive maps: The case of a large
gas refinery
S.M. Asadzadeh a, A. Azadeh a, *, A. Negahban b, A. Sotoudeh a
a
School of Industrial and Systems Engineering, Center of Excellence for Intelligent Based Experimental Mechanics, College of Engineering, University of
Tehran, Tehran, Iran
b
Department of Industrial and Systems Engineering, Auburn University, AL 36849, USA

a r t i c l e i n f o a b s t r a c t

Article history: This study analyzes and assesses the integrated health, safety, environment (HSE) and ergonomics (HSEE)
Received 3 April 2012 factors by fuzzy cognitive maps (FCM) approach. This is achieved through integrating ergonomic and
Received in revised form macro-ergonomic as well as occupational health and safety arrangements in an integrated modeling for
9 January 2013
assessment of their multi-faceted impact on workers’ productivity, injury rate and satisfaction. This
Accepted 16 March 2013
paper uses FCM to assess the direct and indirect effects of HSEE factors on system performance in-
dicators. The results of FCM are used to develop leading indicators useful for proactive management of
Keywords:
productivity, injury rate, and job satisfaction. The result of a comprehensive survey of 37 experts in
Health
Safety
control rooms and maintenance activities in a large gas refinery is used to show the applicability and
Environment (HSE) usefulness of FCM approach. Moreover, FCM results are used to determine the causal structure of HSEE
Macro-ergonomics factors and system performance indicators. It is concluded that macro-ergonomics factors such as in-
Fuzzy cognitive maps (FCM) structions and education, familiarity with organization’s rules, and proper communications most
Gas refinery contribute to improve workers’ safety, satisfaction, and productivity.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction injuries and gets the most efficiency (Changchit & Holsapple, 2001;
Chen & Yang, 2004; Levine & Toffel, 2010).
Occupational health, safety, environment (HSE) management Accidents usually cost a lot of money not only in damage to the
systems at the operational level will strive to eliminate injuries, plant and claims for injury, but also the lost of the company
adverse health effects and damages to the environment. By reputation, workers’ satisfaction and productivity. Accidents and
considering health, safety, environment and ergonomics (HSEE), an occupational ill-health are global phenomena and occur at any
organization manages to regulate, standardize and optimize its time. However, accidents in general and especially those with
operations in a manner that places safety and health first. It en- catastrophic impacts are not unpredictable. Accidents occur prin-
courages employees to adopt a healthy and safe life-style. It de- cipally from interaction between system components arrangement
velops and operates its facilities with due concern for the health i.e. human, technology and organization. The identification of and
and safety of its neighbors and collaborates with authorities in the preparedness for these accidents would either prevent their
preparation of emergency response plans. It contributes to eco- occurrence or mitigate the impact in the case that they do occur.
efficiency by continuously improving energy consumption and The first step in identification and prediction of accidents is to have
reducing waste, emissions and discharges. It designs and develops a model of system components failures and errors. with respect to
products to have the minimum adverse effect on the environment modeling human errors, in the shift from the so-called first gen-
throughout their life-cycle. It optimizes the relation between man eration HRA methods, for example THERP (Swain & Guttmann,
and machine in a manner that man faces the least fatigues, less 1983) to the second generation HRA methods such as ATHEANA
(Cooper et al., 1996) and CREAM (Hollnagel, 1998), the focus of
human failure analysis had changed from inherent deficiencies of
* Corresponding author. humans and task characteristics to the contextual conditions or
E-mail address: aazadeh@ut.ac.ir (A. Azadeh). environment in which the task is performed (Zio, 2009). Therefore,

0950-4230/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jlp.2013.03.007
1016 S.M. Asadzadeh et al. / Journal of Loss Prevention in the Process Industries 26 (2013) 1015e1026

work context and environment are the key factors in analyzing health, environment and safety in Norwegian enterprises. Internal
performance of human workers in the work places. control involves systematic actions that reduce stress and occupa-
Basically, ergonomics is concerned with all those factors that can tional hazards which will, in turn, prevent injuries and workplace
affect people and their behavior (Azadeh, Fam, Khoshnoud, & absenteeism. Flanagan, Strauss, and Ulman (1974) analyzed the
Nikafrouz, 2008). In this sense, ergonomics covers a broad range impacts of job dissatisfaction on work place behavior, such as lower
of programs from work place design, equipment design to work productivity, higher level of strikes, quits and absenteeism. Gyekye
regulation, communication, organization and system design. To and Salminen (2010) examined the relationships between work
better reference to this broad variety of programs, macro- experience and safety perceptions, job satisfaction, compliance
ergonomics refers to the programs and tasks which concerns with safety management policies and accident frequency for Gha-
macro-level in the organization. Training, education, communica- naian industrial workers. Although they did not report causal
tion, regulation and team-work are among those programs and relationship, they showed that there is a positive correlation be-
tasks at macro-level. In micro-level, ergonomics more concerns tween best safety perception, the highest level of job satisfaction,
with tool design, work-place design, lifting and transportation, the most compliant with safety procedures and the lowest accident
work load design, work shift schedule, etc. frequency.
There are close relationship between health, safety, environ- Using ergonomics methods such as subjective assessment,
ment and ergonomics factors (Azadeh, Rouzbahman, Saberi, & direct observation, use of archival data and assessment of noise,
Fam, 2011). Inappropriate design between human and machine Bin, Richardson, and Yeow (2010) studied a conducted ergonomics
could lead to decreased safety. An ergonomics program aims at program including installing noise insulating covers, providing
reducing accidents and ill health, and increase productivity by earplugs, installing elevated platforms, slanting visual display ter-
fitting the task to the worker (Health and Safety Executive). In minals and installing extra exhaust fans. This study showed that the
designing of production systems, economic and social goals can be ergonomics interventions significantly improved workers’ occu-
combined, if ergonomics is integrated into the design process (Dul, pational health and safety, which directly correlated with an
de Vries, Verschoof, Eveleens, & Feilzer, 2004). At organization improvement in working conditions and job satisfaction. Shikdar
level, poor management, training, education, communication, and Sawaqed (2003) identified factors that affected worker pro-
regulation and team-work as well as work environment injurious ductivity, occupational health and safety in industry. Hot environ-
factors could cause human error and safety issues which conse- mental conditions, noisy environment, and lack of resources and
quently would result in environmental risks (Bertolini, 2007; facilities were identified as the main issues. They reported a sig-
Toriizuka, 2001). nificant correlation among productivity indicators and health and
The integrated HSEE system introduces a systemic mechanism, organizational attributes. Lack of skills in ergonomics and training,
which integrates the structure of the human and organizational communication and resources are believed to be some of the fac-
systems with a conventional HSE system and is utilized to enhance tors contributing to the poor ergonomics conditions and conse-
team-work, reliability, availability, maintainability and safety quent loss of worker productivity and reduced health and safety in
(Azadeh, Fam, & Azadeh, 2009; Azadeh Fam, Nouri, & Azadeh, 2008; these industries.
Azadeh, Rouzbahman, & Saberi, 2010). This study attempts to put Past injury/illness metrics as criteria for rating the effectiveness
together ergonomics and macro-ergonomics as well as occupa- of occupational safety and health (OSH) programs are necessary to
tional health and safety arrangements in an integrated modeling assess program success. However, they may not be sufficient for
for assessment of their multi-faceted impact on workers’ produc- developing proactive safety, ergonomics, and environmental
tivity, injury rate and satisfaction. management plans (Wurzelbacher & Jin, 2011). Furthermore, the
Effective application of ergonomics in work system design can effectiveness of occupational safety and health programs needs to
achieve a balance between human characteristics and task de- be assessed in a more broaden context as these programs have an
mands. This can enhance human productivity, reduce human error, influential effect on other worker performance indicators such as
provide improved safety (physical and mental) and job satisfac- productivity and job satisfaction.
tion. Several studies have shown positive effects of applying er- Adopting a holistic view of system performance, this paper uses
gonomics principles to the workplace including machine, job and fuzzy cognitive maps (FCM) to assess the direct and indirect effects
environmental design (Ayoub, 1990a,b; Shikdar & Sawaqed, 2004). of HSEE factors on three system performance indices: workers
Studies in ergonomics have also produced data and guidelines for productivity, injury rate, and job satisfaction. The results of FCM can
industrial applications (Burri & Helander, 1991). However, there is be used to develop leading indicators useful for proactive man-
still a low level of acceptance and limited application in industry. agement of productivity, injury rate, and job satisfaction with re-
The main concern of work system design in context of ergonomics gard to the factors derived from the assessment of HSEE system.
is improvement of machines and tools. Lack of utilization of the Leading indicators of safety performance are predictors of future
ergonomics principles could bring inefficiency to the workplace. safety performance based on selected criteria. The advantage of
Moreover, an ergonomically deficient workplace can cause phys- using leading indicators of safety performance is if performance
ical and emotional stress, low productivity and poor quality of indicators show poor or unacceptable performance, modifications
work conditions. It is believed that ergonomics deficiencies in in- or changes can be made before injuries actually occur.
dustry are root cause of workplace health hazards, low levels of Fuzzy Cognitive Maps have been used in many application
safety and reduced workers’ productivity (Champoux & Brun, fields where decision making process is related with uncertainty,
2003). Although ergonomics applications have gained significant vague and incomplete information (Azadeh, Ziaei, & Moghaddam,
momentum in developed countries, awareness remains low in 2012; Ghaderi, Azadeh, Pourvalikhan Nokhandan, & Fathi, 2012).
developing regions. Ruan and Mkrtchyan (2012) used Belief Degree Fuzzy Cognitive
Considerable research works have been devoted to study the Maps to find the main factors influencing strong safety culture.
interrelationships of occupational health and safety, workers pro- Safety Culture has been described how safety issues are managed
ductivity and job satisfaction. Azadeh et al. (2011) based on adap- within an enterprise. Chytas, Glykas, and Valiris (2010) described
tive neural network (ANN) algorithm analyzed the relationship an FCM model for software reliability prediction. In order for
between operators job satisfaction and HSEE factors. Saksvik and dealing with vague information and taking into account human
Nytr (1996) presented an implementation of internal control of common sense and intuition, Smith and Eloff (2000) proposed a
S.M. Asadzadeh et al. / Journal of Loss Prevention in the Process Industries 26 (2013) 1015e1026 1017

fuzzy cognitive maps approach to assess risk in health care surveys. These decisions may result in either positive or dramatic
institutions. consequences that highly depend on the correct evaluation deci-
Toriizuka (2001) collected the background factors of human sional factors. Therefore, decision makers need to know all factors
behavior in an industrial plant and proposed original performance related to the problem in order to make decisions. FCMs provide
shaping factors (PSFs) for maintenance tasks and utilized them to decision makers with a tool to support and make successful
evaluate human reliability. He categorized PSFs into eight groups: decisions.
1) Judgmental load, 2) Physical load, 3) Mental load, 4) Information In fact, a Fuzzy Cognitive Map incorporates the accumulated
and confirmation, 5) Indication and communication, 6) Machinery experience and knowledge by employing human experts who are
or tools, 7) Environment, 8) Work space. The PSFs defined by aware of system operations and behavior in different situations.
Toriizuka (2001) were considered by Bertolini (2007) and a fuzzy The main advantage of this method is its simplicity. Another
cognitive maps (FCMs) approach was presented in order to explore advantage is that it functions on expert’s opinion. Experts deter-
the importance of the factors affecting human reliability in indus- mine concepts, interconnections, and assign casual fuzzy weights
trial plants. The two macro-categories ‘‘environment’’ and ‘‘indi- to the interconnections. However, the strength of the data de-
cation and communication’’ have emerged as the most important pends on the number of experts’ opinion we can use. FCM is the
categories as regards the causal influence on the top event human only fuzzy technique that gives the hidden pattern of the
reliability. problem.
With the current increasing need for safety, productivity and Decision makers face difficulties in complicated dynamic sys-
efficiency of both plant and human operator, fuzzy cognitive maps tems where mathematical model formulation can be difficult,
(FCM) will provide an organization with a valid help in assessing costly, and impossible in some cases. In general, FCMs have been
the most critical factors in managing socio-technical systems found useful in variety of applications, including administrative
(Bertolini & Bevilacqua, 2010). FCM of this study is a means to study sciences, game theory, information analysis, popular political de-
HSEE system factors and obtain useful indications on the conse- velopments, electrical circuit analysis, cooperative manemachines,
quences which can be determined by the variation of one or more distributed group-decision support, adaptation and learning
health, safety, environment, ergonomics, and macro-ergonomics (Craiger, Goodman, Wiss, & Butler, 1996; Dickerson & Kosko, 1994;
variables. They can provide an interesting solution to the issue of Ghaderi et al., 2012).
assessing the factors which are considered to affect the operator’s Fuzzy Cognitive Maps describe the behavior of a complex sys-
safety, productivity and satisfaction. tem in terms of concepts which represent different states, charac-
The remainder of the paper is organized as follows: Section 2 teristics, or variables of the system. Concepts interact with each
presents the overall methodology of the paper and illustrates the other in order to show the dynamics of the system. FCMs illustrate
method for analyzing cyclic FCMs. The research method of the the whole system by a graph showing the cause and effect relations
paper is discussed in section 3. Section 4 shows the application of that exists among these concepts. The Fuzzy Cognitive Maps were
FCM for an HSEE system in gas refineries. Last section summarizes first introduced by Kosko (1986). He enhanced the power of
and ends with the main findings of the study. cognitive maps (Axelord, 1976), by considering fuzzy linguistic
representation for the concepts and fuzzy degrees of in-
2. The overall methodology terrelationships between concepts.
FCM is a directed graph which is composed of concepts (nodes)
The methodology of the paper is designed so as it comes up with and causal connections (edges) in order to represent causal rela-
constructed indices of system performance in the integrated HSEE tionship between concepts. The degree of the relationship between
system. The overall methodology of the paper is presented in six concepts can be described as a number between [1, 1], or fuzzy
stages as follows: linguistic terms, such as ‘‘high’’, ‘‘low’’, ‘‘medium’’, etc. Besides, each
variable concept can be excited in different measure from 0% to
Stage 1. Identification of HSEE system limits 100%.
Stage 2. Identification of system performance indicators The following rules are proposed by (Kardaras & Karakostas,
Stage 3. Elicitation of concepts: Experts elicit a set of factors 1999) in order to analyze cognitive maps:
(here called concepts) that might have an effect on system
performance  If an increase or decrease in Concept 1 causes Concept 2 to
Stage 4. Direct impacts: the direct impact of each concept on change in the same direction, the relationship is positive. If the
system performance as well as the impacts on each other will be effect on the effect variable is in the opposite direction, the
determined through subjective assessment. relationship is negative.
Stage 5. Overall concepts weights: A weight of concept is the  The indirect effect I(X,Y) of a cause variable (X) on an effect
relative importance of that concept in comparison to the variable (Y) is positive if the multiplication of the sign of causal
concept judged to be the most important. Fuzzy cognitive maps effects that form the path from the cause variable to the effect
is used to calculate these weights variable is positive.
Stage 6. Construct leading indicators of system performance e  The total effect of a cause variable X on effect variable Y is the
based on the weight results of stage 5, indicators of system sum of all the direct and indirect effects from all the possible
performance are constructed for the analysis of changes in HSEE paths that connect X to Y and is denoted by T(X, Y).
factors.
Cyclic FCM includes at least a directed cycle (See Fig. 1 as an
2.1. Introduction to FCM example) whereas acyclic FCM does not have any directed cycle.
The interest for analysis of cyclic FCMs is their capability to model
Fuzzy Cognitive Map (FCM) methodology is a symbolic repre- dynamic systems with feedback. The counterpart of feedbacks in
sentation that is used for describing and modeling complex systems. FCM is a cycle.
FCMs provide excellent mechanisms for supporting decision-makers In the cyclic FCM of Fig. 1, suppose we are interested in calcu-
(Kosko, 1986). Decisions in organizations are commonly complex lating the total effects of C1 on C4. According to Kosko (1986), the
which require analysis, brainstorming, experts’ consultation, or total effects are defined as follows. Concept 1 is connected to
1018 S.M. Asadzadeh et al. / Journal of Loss Prevention in the Process Industries 26 (2013) 1015e1026

a lot 2.2. Applied FCM procedure


C1 C2
A general algorithm for performing FCM is given as follows:
- some
some - a lot - some Step 1: Read the input vector A(t).
Step 2: Give the connection matrix, E.
Step 3: Calculate the output vector O(t) ¼ A(t) *E.
C3 a lot C4 Step 4: Apply threshold function f($) to output vector: O(t) ¼ f
{O(t)}.
Fig. 1. Example of cyclic FCM. Step 5: If (O(t) ¼ A(t)), stop
else set t ¼ t þ 1; A(t) ¼ O(t); go to step 3.
End
Concept 4 via an infinite number of paths because the FCM in Fig. 1
is a cyclic graph with one loop between concepts C2 and C4. A In the Step 4 of the general algorithm, f is a threshold function
fuzzy causal algebra is developed by Kosko (1986) to effectively which can be defined in various formulations. The type of threshold
govern the causal propagation and combinations in FCM. Here the function has significant impact on the result of FCM analysis. Two
Min-Max fuzzy algebraic operators are illustrated. There is one threshold functions are usually used: The uni-polar sigmoid func-
direct effect of C1 on C4 which is effective on the one-edge Path 1: tion when interactions can only put on positive numbers and
P1(C1eC4). The effect delivered on this path is I1(1,4) and calcu- tanh(x) when concepts can be negative and their values belong to
lated as the interval [1, 1]. Thresholding usually helps quickly converging
to stable limit cycles or fixed points. These limit cycles show
/I1 ð1; 4Þ ¼ minfsomeg ¼ some; “hidden patterns” in the causal web of FCM (Kosko, 1988).
Here, to analyze a cyclic cognitive map we propose to use a
Indirect effect is an effect that a concept has on another concept normalizing function as the threshold function f. What makes this
via other concepts in FCM. The first group of indirect effects of C1 on choice appropriate is that in the FCM analysis of this study we are
C4 consists of two two-edge paths. C1 has an indirect effect on C4 interested in the relative total effects of concepts. Specifically by
via C2. Two-edge Path 2: P2(C1-C2eC4) represents this effect. having the relative total effects of HSEE factors on system perfor-
mance indicators such as injury rate, we are able to develop leading
indicators to improved managing the system performance. In this
/I2 ð1; 4Þ ¼ minfa lot;  a lotg ¼ a lot; study for the analysis of a cyclic cognitive map, the FCM matrix
properties showed by Kosko (1986) are used and a Monte Carlo
C1 has also an indirect effect on C4 via C3. Two-edge Path 3:
simulation-based procedure is suggested as follows:
P3(C1eC3eC4) represents this effect.

/I3 ð1; 4Þ ¼ minfsome; a lotg ¼ some: Table 1


HSEE system and its performance concepts.
Because of the loop (C2eC4) in the FCM, C4 could have indirect
Category Symbol Concepts
effect on itself via C2. Therefore, the second group of indirect effects
Health H1 Lack of enough drinks and liquids in warm
consists of one three-edge path P4(C1eC4eC2eC4) from C1 to C4
places
including a loop. H2 Lack of fresh air
H3 Breathing problems in workplace
H4 Stress and mental pressures
/I4 ð1; 4Þ ¼ minfsome;  some;  a lotg ¼ some: Safety S1 Not using safety devices and procedures
S2 Uncomfortable and obstructive safety devices
This means that C4 have an effect on itself via C2 because once S3 Dangerous and unsuitable work space design
C4 is affected by C1, it will affect C2 and once C2 is affected, it S4 Hazards in work place
S5 Fire and explosion hazards in work place
again affects C4. This process continually happens until the total
Environment EN1 Loud noise
effect reaches equilibrium. Accordingly, there are an infinite EN2 Improper lighting
number of indirect effects this FCM. Similarly four-edge paths, EN3 Warm places
five-edge paths, etc can be drawn and their indirect effects can be EN4 Pollution in the workplace
calculated. EN5 Pollution nearby workplace
Ergonomy ER1 Pain and distress because of work
The total effect of C1 on C4 is calculated by applying the max ER2 Ergonomically unsuitable office furniture
operator on all direct and indirect paths from C1 to C4. Due to the ER3 Difficult to move within workplace
presence of loops in the three or more-edge paths, the number of ER4 Difficult to transport and operate machines,
indirect effects is infinite and there is no closed form solution for materials and tools
ER5 Vibrations of tools and machinery on body
the total effects. Therefore, the equilibrium state for such dy-
ER6 Ergonomically unsuitable machines and tools
namic systems is of interest. Moreover, for the numerical analysis Macro ergonomy MER1 Lack of instructions and education about safety
of a dynamic system usually fuzzy numbers are defuzzified into and accident prevention
crisp numbers and by doing so, FCMs can numerically be MER2 Lack of respect for rules
analyzed. Acyclic FCMs can be numerically analyzed using the MER3 Work pressures
MER4 Lack of documented instructions about works
traditional approach proposed by Kosko (1986) but they consid-
MER5 Unfamiliarity with organization’s rules
ered FCM as a directed acyclic graph with no causal loops. MER6 Lack of proper communications with managers
Violation from the assumption of FCM as directed acyclic graph MER7 Difficulties within organization with coworkers
however restricts the application of the traditional procedure. System performance SP1 Overall job satisfaction
SP2 Workers productivity
The next section presents a procedure for numerical analysis of
SP3 Injury rates
cyclic FCMs.
S.M. Asadzadeh et al. / Journal of Loss Prevention in the Process Industries 26 (2013) 1015e1026 1019

HSEE Categories System Performance

Macro-
Health
Ergonomy
Job Workers
Satisfaction Productivity

Safety Ergonomy

Injury Rate

Environment

Fig. 2. The research framework.

Let C1, C2, ., Cn be causal concepts and let eij ¼ e(Ci, Cj) be the Table 2
Cronbach alpha for the different categories.
(i, j) cell in adjacency matrix E representing the amount of cau-
sality Ci imparts to Cj. It lists all the one-edge paths on the Category Health Safety Environment Ergonomy Macro ergonomy
cognitive map representing the direct effects of concepts. ra 0.71 0.82 0.77 0.72 0.83
E2 ¼ ½e2ij  ¼ E*E lists all two-edge paths on the cognitive map.
Similarly E3, E4, ., En1list the effect of summing all three-edge,
four-edge,., and (n1)-edge indirect effects. The total effect of
Ci on Cj is the summation of direct and all indirect effects. In an 3. Research method
acyclic FCM, indirect effects corresponding to edges with more
than (n1) order all are zero. However, here E is not an acyclic 3.1. Concept development for FCM
matrix and the indirect effects include high-order edge effects.
Therefore, En, Enþ1, ., EM (M is a big number approaching infinity) The integrated HSEE system characteristics as well as its per-
all include in total indirect effects. Then total-effect matrix TM can formance, give shape to five categories of concepts in FCM: Health,
be calculated by Equation (1). safety, environment, ergonomics, macro ergonomics, and system
performance. Initially, a literature review was carried out, consid-
XM/N ering the studies on occupational safety and health, ergonomics,
TM ¼ i¼1
Ei (1) customer satisfaction and productivity in industrial plants specif-
ically for gas refineries. Standard indicators for assessment of HSEE
It is noted that each row of TM contains the total effects of one factors developed by Azadeh, Fam, Khoshnoud, et al. (2008) are
concept on other concepts in FCM and each column of TM contains used for system quantification in each category. For validity anal-
the total influence by all the concepts in FCM on a single concept. ysis, the method of content validity is used. These factors are
Furthermore, a normalized column in TM represents the relative initially analyzed by a panel of 7 experts to verify the relevance of
causal weight of other concepts in forming the concept corre- concepts for HSEE assessment. Based on this initial validation, the
spondent to that column. most important factors (concepts in FCM terminology) are finalized
Although for a cyclic cognitive map TM does not necessarily and represented in Table 1.
represent a deterministic stable process with known steady state; Fig. 2 illustrates the research framework describing the rela-
however the normalized matrix Norm (TM) resulted from tionship between the HSEE categories (independent variables) and
normalizing columns of TM by dividing them by the maximum cell system performance indicators (dependent variables). Each HSEE
of TM shows a stable behavior. This can be examined easily by category represents a set of system aspects which are hypothesized
Monte Carlo simulation of TM process. The following simple to affect other aspects. Furthermore, the categories are interrelated
pseudo code illuminates the process of Monte Carlo simulation in and are assumed to have impact on each other and on system
FCM. performance. The proposed FCM of the study analyses the total
1020
Table 3
The global FCM connection matrix.

S.M. Asadzadeh et al. / Journal of Loss Prevention in the Process Industries 26 (2013) 1015e1026
H1 H2 H3 H4 S1 S2 S3 S4 S5 EN1 EN2 EN3 EN4 EN5 ER1 ER2 ER3 ER4 ER5 ER6 MER1 MER2 MER3 MER4 MER5 MER6 MER7 SP1 SP2 SP3
H1 0 0 0 0.64 0.38 0 0 0 0 0 0 0 0 0 0.23 0 0 0 0 0 0 0.24 0.42 0 0 0.37 0.37 0.74 0.72 0.33
H2 0 0 0.89 0.56 0.51 0 0 0 0 0 0.49 0.27 0.42 0.4 0.55 0 0 0 0 0 0 0.36 0.53 0 0 0.22 0.24 0.83 0.66 0.51
H3 0 0 0 0.7 0.45 0 0 0.38 0.25 0 0 0 0 0 0.59 0 0 0 0 0 0 0.31 0.65 0 0 0.22 0.24 0.69 0.71 0.73
H4 0 0 0 0 0.32 0 0 0.56 0.46 0 0 0 0 0 0.52 0 0 0 0 0 0 0.51 0.7 0 0 0.53 0.69 0.68 0.68 0.54
S1 0 0.14 0.59 0.34 0 0 0 0.75 0.6 0 0 0 0 0 0.47 0 0 0 0.65 0 0 0.23 0 0 0 0 0 0.39 0.52 0.79
S2 0 0 0.56 0.66 0.82 0 0 0.7 0.55 0 0 0 0 0 0.72 0 0.38 0 0.54 0 0 0.45 0.44 0 0 0 0 0.63 0.74 0.49
S3 0 0 0.42 0.5 0.63 0 0 0.74 0.71 0.23 0.61 0.39 0.35 0.21 0.63 0 0.68 0.68 0.32 0 0 0.44 0.53 0 0 0.31 0.43 0.59 0.76 0.65
S4 0 0 0 0.56 0.58 0 0 0 0.58 0 0 0 0 0 0.33 0 0 0 0 0 0.41 0.33 0.59 0.46 0.2 0.18 0.34 0.49 0.32 0.74
S5 0 0 0 0.5 0.59 0 0 0.66 0 0 0 0 0 0 0.47 0 0 0 0 0 0.53 0.34 0.55 0.69 0.21 0.2 0.19 0.58 0.36 0.84
EN1 0 0 0 0.42 0.5 0 0 0.46 0.22 0 0 0 0 0 0.57 0 0 0 0 0 0 0.27 0.67 0 0 0.28 0.43 0.56 0.69 0.46
EN2 0 0 0 0.49 0.33 0 0 0.6 0.46 0 0 0 0 0 0.51 0 0.53 0.72 0 0 0 0.37 0.55 0 0 0.24 0.3 0.54 0.78 0.63
EN3 0.54 0 0.51 0.66 0.56 0 0 0.48 0.64 0 0 0 0.35 0.19 0.65 0 0 0.53 0 0 0 0.44 0.65 0 0 0.41 0.39 0.74 0.79 0.51
EN4 0 0.71 0.75 0.66 0.53 0 0 0.5 0.48 0 0.4 0 0 0.53 0.47 0 0 0 0 0 0 0.36 0.54 0 0 0.5 0.5 0.64 0.71 0.63
EN5 0 0.61 0.63 0.38 0.29 0 0 0.26 0.24 0 0.31 0 0.54 0 0.38 0 0 0 0 0 0 0 0.23 0 0 0 0 0.35 0.23 0.53
ER1 0 0 0 0.67 0.45 0 0 0.57 0.34 0 0 0 0 0 0 0 0 0 0 0 0 0.41 0.57 0 0 0.41 0.47 0.88 0.88 0.71
ER2 0 0 0 0.38 0.28 0 0.41 0.33 0.23 0 0.47 0 0 0 0.57 0 0.51 0.33 0.15 0 0 0 0 0 0 0 0 0.55 0.58 0.32
ER3 0 0 0 0.43 0.55 0.42 0 0.59 0.39 0 0 0 0 0 0.48 0 0 0.6 0 0 0 0.37 0.55 0 0 0.44 0.43 0.46 0.65 0.44
ER4 0 0 0 0.3 0.45 0.44 0 0.69 0.62 0.23 0 0 0.37 0.23 0.63 0 0.74 0 0.68 0 0 0.35 0.61 0 0 0 0 0.45 0.67 0.6
ER5 0 0 0 0.45 0.45 0 0 0.54 0.46 0 0 0 0 0 0.77 0 0 0.66 0 0 0.39 0 0.61 0 0 0 0 0.66 0.56 0.79
ER6 0 0 0 0.42 0.46 0.49 0 0.66 0.62 0.48 0 0 0 0 0.81 0 0.4 0.74 0.79 0 0 0.46 0.79 0 0 0.36 0.35 0.53 0.84 0.69
MER1 0 0 0 0.35 0.72 0.62 0.61 0.84 0.82 0 0.46 0.44 0.38 0.22 0.6 0 0.21 0.31 0.66 0.49 0 0.5 0.71 0 0.62 0.35 0.31 0.45 0.66 0.83
MER2 0 0 0 0.26 0.76 0 0.55 0.81 0.75 0.2 0.39 0.25 0.42 0.36 0.42 0 0 0 0 0 0.53 0 0.2 0.5 0.51 0.38 0.4 0 0.59 0.64
MER3 0 0 0 0.69 0.39 0 0 0.48 0.51 0 0 0 0 0 0.78 0 0 0 0 0 0 0.52 0 0 0 0.39 0.64 0.83 0.61 0.58
MER4 0 0.32 0 0.32 0.71 0 0.51 0.74 0.74 0.34 0.34 0.38 0.5 0.39 0.61 0.16 0.28 0.57 0.68 0.44 0.83 0.7 0.57 0 0.62 0.29 0.5 0.46 0.73 0.63
MER5 0 0 0 0.34 0.51 0 0.48 0.48 0.51 0.27 0.22 0.21 0.43 0.35 0.43 0 0.33 0.35 0 0.38 0.53 0.85 0.52 0 0 0.56 0.37 0.27 0.54 0.53
MER6 0.4 0 0 0.51 0 0.39 0.49 0.38 0.39 0.19 0.43 0.36 0.41 0.25 0.29 0.34 0.43 0.53 0 0.47 0.76 0.46 0.4 0.73 0.55 0 0.35 0.51 0.56 0.23
MER7 0 0 0 0.56 0 0 0 0.28 0.28 0 0 0 0 0 0.28 0 0 0 0 0 0.39 0.27 0.34 0 0.52 0.52 0 0.65 0.54 0.32
SP1 0 0 0 0.65 0.48 0 0 0.33 0.25 0 0 0 0.22 0.22 0.44 0 0 0 0 0 0 0.71 0.52 0 0.41 0.35 0.68 0 0.8 0.28
SP2 0 0 0 0.41 0.27 0 0 0.37 0.25 0 0 0 0.35 0.24 0 0 0 0 0 0 0.23 0.25 0.49 0.32 0.32 0.34 0.34 0.55 0 0.26
SP3 0 0 0 0.72 0.68 0 0.52 0.65 0.45 0 0 0 0 0 0.76 0 0 0 0 0 0.47 0.66 0.51 0.59 0.33 0.44 0.45 0.87 0.86 0
S.M. Asadzadeh et al. / Journal of Loss Prevention in the Process Industries 26 (2013) 1015e1026 1021

effects of HSEE variables and determines the relative significance of x 10


20
10
the existing effects.
8
3.2. Causal interrelationships
6
A panel of 37 experts consisting plant managers and academic

Total effect
researchers was formed to specify the causal interrelationships of 4
the concepts. Basically, the target experts have a high level of
background in system performance analysis and human resource 2
management. With background we mean executive long-term
experience as well as expertise and knowledge about health and 0
safety systems, employee relationship management, and human
productivity management. Academic researchers in the panel have -2
been long studying the related subjects and have published
research papers in the fields of system performance analysis and -4
0 5 10 15 20 25 30
human resource management. M

Respondents are asked to express their opinion about the paired


Fig. 3. Changes in the total effects of HSEE factors on injury rate (SP3).
relationships of concepts, based on a five-point Likert-type scale
with anchors ranging from ‘strongly negative (1)’ to ‘strongly
positive (þ1)’ including ‘positive (þ0.5)’, ‘neutral (0)’, and ‘negative where FCMglobal is the global FCM that represents the expert
(0.5)’. Respondents realize that a value of zero means no causal domain knowledge, Credi is the credibility weight of expert i, and
influence between concepts. A positive value of relationship shows FCMi is the single FCM of expert i. Considering equal credibility for
that an increase in one concept leads to an increase in the other experts, the global FCM is obtained as inTable 3. Although the in-
concept and a negative value of interrelationship shows an increase dividual FCMi represents a fuzzy cognitive map, the global FCM can
in one concept cause the other concept to decrease. be considered as a cognitive map since the uncertainty arise from
Reliability is a technical specification of the instrument and specifying the causal relationships by only a single expert has been
shows the ability of the instrument to generate identical results in a handled through aggregation of all expert opinions. This global
given circumstance. Reliability is a coefficient between zero and one FCM is considered as the adjacency matrix for the cognitive map
and indicates the relevance and assessment power of the instru- analysis in the next section.
ment. Here, to examine the reliability of the instrument, the Cron-
bach alpha is used. The Cronbach alpha measures the internal 4. Results
coordination of the instrument and is calculated by the Equation (2).
In order for the analysis of global FCM, the suggested procedure
0 1 of Section 2.2 is used. Suppose the FCM is a cognitive map, then the
Pn global FCM represented in Table 3 is an adjacency matrix (E). It lists
J B
2C
B j ¼ 1 sj C
ra ¼ B1  C (2) all the one-edge paths on the cognitive map representing the direct
J 1@ S 2 A effects of concepts. Then for total-effect matrix TM for which up to
M-edge paths have been aggregated, Equation (1) is used. It is noted
that each row of TM contains the total effects of one concept on
Where J is the number of concepts in each category which is other concepts in FCM and each column of TM contains the total
identical to the number of questions in the questionnaire. s2j is the influence by all the concepts in FCM on a single concept. Norm (TM)
variance of jth concept, and S2 is the overall variance. As com- normalizes the columns of TM by dividing them by the maximum of
mented by Dornyei (2007, p. 207), the accepted level of this statistic TM. Furthermore, a normalized column in TM represents the relative
is 0.7. In the other words, ra greater than 0.7 shows acceptable causal weight of other concepts in forming the concept corre-
reliability of the instrument. The ra of the questionnaire in different spondent to that column.
categories is calculated by SPSS 18 software and results are pre-
sented in Table 2. The results show an acceptable reliability level for 0.5
the questionnaire in different categories.
The signed FCM can show the cause and effect relations be-
0.4
tween factors; however in order to numerically and quantitatively
analyze the system, we need to give weights to relationships. In
Normalized total effect

0.3
order to do so, there are different approaches (Papageorgiou,
Stylios, Groumpos, & Peter, 2003; Stach, Kurgan, Pedrycz, &
0.2
Reformat, 2005; Taber, 1991). In one approach, experts’ opinions
in the form of linguistic terms are gathered and after combining all
0.1
the different experts’ opinions together and then using deffuzifying
techniques such as center of area (COA) a weight is assigned to each
0
relationship. Here, the formula suggested by Taber (1991) is used to
unify different judgments of experts, which is based on the credi-
-0.1
bility weights of the experts. The union of different judgments is
obtained through Equation (3):
-0.2
0 5 10 15 20 25 30
XN M
FCMglobal ¼ i¼1
ðFCMi *Credi Þ (3)
Fig. 4. Changes in the normalized total effects on injury rate of HSEE factors.
1022
Table 4
FCM normalized total effect matrix.

S.M. Asadzadeh et al. / Journal of Loss Prevention in the Process Industries 26 (2013) 1015e1026
H1 H2 H3 H4 S1 S2 S3 S4 S5 EN1 EN2 EN3 EN4 EN5 ER1 ER2 ER3 ER4 ER5 ER6 MER1 MER2 MER3 MER4 MER5 MER6 MER7 SP1 SP2 SP3
H1 0.01 0.02 0.03 0.31 0.01 0.02 0.01 0.22 0.19 0.02 0.04 0.02 0.08 0.06 0.26 0.01 0.03 0.03 0.01 0.02 0.01 0.14 0.28 0.02 0.07 0.14 0.26 0.36 0.36 0.29
H2 0.02 0.03 0.04 0.48 0.02 0.03 0.01 0.33 0.29 0.03 0.06 0.03 0.12 0.10 0.40 0.01 0.05 0.05 0.02 0.03 0.02 0.22 0.44 0.03 0.11 0.22 0.40 0.55 0.56 0.45
H3 0.01 0.02 0.03 0.30 0.01 0.02 0.01 0.21 0.18 0.02 0.04 0.02 0.08 0.06 0.25 0.01 0.03 0.03 0.01 0.02 0.01 0.14 0.28 0.02 0.07 0.14 0.25 0.35 0.35 0.28
H4 0.01 0.02 0.03 0.34 0.01 0.02 0.01 0.24 0.21 0.02 0.04 0.02 0.09 0.07 0.29 0.01 0.03 0.04 0.01 0.02 0.01 0.16 0.31 0.02 0.08 0.16 0.29 0.39 0.40 0.32
S1 0.01 0.01 0.02 0.16 0.01 0.01 0.00 0.11 0.10 0.01 0.02 0.01 0.04 0.03 0.14 0.00 0.02 0.02 0.01 0.01 0.01 0.08 0.15 0.01 0.04 0.08 0.14 0.19 0.19 0.16
S2 0.01 0.02 0.03 0.33 0.01 0.02 0.01 0.23 0.20 0.02 0.04 0.02 0.08 0.07 0.28 0.01 0.03 0.04 0.01 0.02 0.01 0.15 0.30 0.02 0.07 0.15 0.28 0.38 0.39 0.31
S3 0.02 0.03 0.05 0.52 0.02 0.03 0.01 0.37 0.32 0.03 0.07 0.04 0.14 0.11 0.44 0.02 0.05 0.06 0.02 0.03 0.02 0.24 0.48 0.04 0.12 0.25 0.45 0.60 0.62 0.50
S4 0.00 0.00 0.01 0.08 0.00 0.01 0.00 0.06 0.05 0.00 0.01 0.01 0.02 0.02 0.07 0.00 0.01 0.01 0.00 0.00 0.00 0.04 0.07 0.01 0.02 0.04 0.07 0.09 0.09 0.08
S5 0.01 0.01 0.01 0.14 0.01 0.01 0.00 0.09 0.08 0.01 0.02 0.01 0.04 0.03 0.11 0.00 0.01 0.01 0.00 0.01 0.01 0.06 0.12 0.01 0.03 0.06 0.11 0.16 0.16 0.13
EN1 0.01 0.02 0.02 0.26 0.01 0.02 0.01 0.18 0.16 0.01 0.03 0.02 0.07 0.06 0.22 0.01 0.02 0.03 0.01 0.02 0.01 0.12 0.24 0.02 0.06 0.12 0.22 0.30 0.31 0.25
EN2 0.01 0.02 0.03 0.33 0.01 0.02 0.01 0.23 0.21 0.02 0.04 0.02 0.09 0.07 0.28 0.01 0.03 0.04 0.01 0.02 0.01 0.15 0.31 0.02 0.08 0.16 0.28 0.39 0.40 0.32
EN3 0.02 0.03 0.05 0.50 0.02 0.03 0.01 0.35 0.31 0.03 0.06 0.03 0.13 0.10 0.42 0.01 0.05 0.06 0.02 0.03 0.02 0.23 0.46 0.03 0.11 0.23 0.42 0.57 0.59 0.47
EN4 0.02 0.03 0.04 0.47 0.02 0.03 0.01 0.33 0.29 0.03 0.06 0.03 0.12 0.10 0.40 0.01 0.04 0.05 0.02 0.03 0.02 0.22 0.43 0.03 0.11 0.22 0.40 0.54 0.56 0.45
EN5 0.01 0.01 0.02 0.24 0.01 0.01 0.00 0.17 0.15 0.01 0.03 0.02 0.06 0.05 0.20 0.01 0.02 0.03 0.01 0.01 0.01 0.11 0.22 0.02 0.05 0.11 0.20 0.28 0.28 0.23
ER1 0.01 0.02 0.03 0.31 0.01 0.02 0.01 0.21 0.19 0.02 0.04 0.02 0.08 0.06 0.26 0.01 0.03 0.03 0.01 0.02 0.01 0.14 0.28 0.02 0.07 0.14 0.26 0.35 0.36 0.29
ER2 0.01 0.02 0.03 0.27 0.01 0.02 0.01 0.19 0.17 0.01 0.03 0.02 0.07 0.06 0.23 0.01 0.03 0.03 0.01 0.02 0.01 0.12 0.25 0.02 0.06 0.13 0.23 0.31 0.32 0.26
ER3 0.02 0.02 0.03 0.37 0.02 0.02 0.01 0.26 0.23 0.02 0.05 0.02 0.10 0.08 0.31 0.01 0.03 0.04 0.01 0.02 0.01 0.17 0.34 0.03 0.08 0.17 0.31 0.43 0.44 0.35
ER4 0.02 0.02 0.03 0.35 0.01 0.02 0.01 0.24 0.21 0.02 0.04 0.02 0.09 0.07 0.29 0.01 0.03 0.04 0.01 0.02 0.01 0.16 0.32 0.02 0.08 0.16 0.29 0.40 0.41 0.33
ER5 0.01 0.01 0.01 0.15 0.01 0.01 0.00 0.11 0.09 0.01 0.02 0.01 0.04 0.03 0.13 0.00 0.01 0.02 0.01 0.01 0.01 0.07 0.14 0.01 0.03 0.07 0.13 0.17 0.18 0.14
ER6 0.02 0.03 0.05 0.49 0.02 0.03 0.01 0.34 0.30 0.03 0.06 0.03 0.13 0.10 0.41 0.01 0.05 0.05 0.02 0.03 0.02 0.22 0.45 0.03 0.11 0.23 0.41 0.56 0.58 0.46
MER1 0.03 0.04 0.06 0.68 0.03 0.04 0.01 0.48 0.42 0.04 0.09 0.05 0.18 0.14 0.57 0.02 0.06 0.08 0.02 0.04 0.03 0.31 0.63 0.05 0.16 0.32 0.58 0.79 0.81 0.65
MER2 0.02 0.03 0.05 0.49 0.02 0.03 0.01 0.34 0.30 0.03 0.06 0.03 0.13 0.10 0.41 0.01 0.05 0.05 0.02 0.03 0.02 0.23 0.45 0.03 0.11 0.23 0.42 0.57 0.58 0.47
MER3 0.01 0.02 0.03 0.34 0.01 0.02 0.01 0.24 0.21 0.02 0.04 0.02 0.09 0.07 0.29 0.01 0.03 0.04 0.01 0.02 0.01 0.16 0.31 0.02 0.08 0.16 0.29 0.39 0.40 0.32
MER4 0.04 0.05 0.08 0.85 0.04 0.05 0.02 0.59 0.52 0.05 0.11 0.06 0.22 0.18 0.71 0.02 0.08 0.09 0.03 0.05 0.03 0.39 0.78 0.06 0.19 0.40 0.72 0.97 1.00 0.80
MER5 0.03 0.04 0.06 0.63 0.03 0.04 0.01 0.44 0.39 0.03 0.08 0.04 0.16 0.13 0.53 0.02 0.06 0.07 0.02 0.04 0.02 0.29 0.58 0.04 0.14 0.30 0.54 0.73 0.75 0.60
MER6 0.04 0.05 0.08 0.84 0.04 0.05 0.02 0.59 0.52 0.05 0.11 0.06 0.22 0.18 0.71 0.02 0.08 0.09 0.03 0.05 0.03 0.39 0.77 0.06 0.19 0.39 0.71 0.97 0.99 0.80
MER7 0.02 0.02 0.03 0.37 0.02 0.02 0.01 0.26 0.23 0.02 0.05 0.02 0.10 0.08 0.31 0.01 0.04 0.04 0.01 0.02 0.01 0.17 0.34 0.03 0.09 0.18 0.32 0.43 0.44 0.35
SP1 0.02 0.02 0.04 0.41 0.02 0.03 0.01 0.29 0.25 0.02 0.05 0.03 0.11 0.09 0.35 0.01 0.04 0.05 0.01 0.02 0.02 0.19 0.38 0.03 0.09 0.19 0.35 0.47 0.49 0.39
SP2 0.02 0.02 0.03 0.36 0.02 0.02 0.01 0.25 0.22 0.02 0.05 0.02 0.09 0.08 0.30 0.01 0.03 0.04 0.01 0.02 0.01 0.17 0.33 0.02 0.08 0.17 0.31 0.42 0.43 0.34
SP3 0.00 0.01 0.01 0.10 0.00 0.01 0.00 0.07 0.06 0.01 0.01 0.01 0.03 0.02 0.08 0.00 0.01 0.01 0.00 0.01 0.00 0.04 0.09 0.01 0.02 0.05 0.08 0.11 0.11 0.09
S.M. Asadzadeh et al. / Journal of Loss Prevention in the Process Industries 26 (2013) 1015e1026 1023

Table 5
Effects and influence of FCM concepts.

Category Symbol Concepts Absolute row mean Absolute column mean


Health H1 Lack of enough drinks and liquids in warm places 0.1 0.0
H2 Lack of fresh air 0.2 0.0
H3 Breathing problems in workplace 0.1 0.0
H4 Stress and mental pressures 0.1 0.4
Safety S1 Not using safety devices and procedures 0.1 0.0
S2 Uncomfortable and obstructive safety devices 0.1 0.0
S3 Dangerous and unsuitable work space design 0.2 0.0
S4 Hazards in work place 0.0 0.3
S5 Fire and explosion hazards in work place 0.0 0.2
Environment EN1 Loud noise 0.1 0.0
EN2 Improper lighting 0.1 0.0
EN3 Warm places 0.2 0.0
EN4 Pollution in the workplace 0.2 0.1
EN5 Pollution nearby workplace 0.1 0.1
Ergonomy ER1 Pain and distress because of work 0.1 0.3
ER2 Ergonomically unsuitable office furniture 0.1 0.0
ER3 Difficult to move within workplace 0.1 0.0
ER4 Difficult to transport and operate machines, materials and tools 0.1 0.0
ER5 Vibrations of tools and machinery on body 0.1 0.0
ER6 Ergonomically unsuitable machines and tools 0.2 0.0
Macro ergonomy MER1 Lack of instructions and education about safety and accident prevention 0.2 0.0
MER2 Lack of respect for rules 0.2 0.2
MER3 Work pressures 0.1 0.4
MER4 Lack of documented instructions about works 0.3 0.0
MER5 Unfamiliarity with organization’s rules 0.2 0.1
MER6 Lack of proper communications with managers 0.3 0.2
MER7 Difficulties within organization with coworkers 0.1 0.3
System performance SP1 Overall job satisfaction 0.1 0.4
SP2 Workers productivity 0.1 0.5
SP3 Injury rates 0.0 0.4

The results of Monte Carlo simulation of both TM and Norm (TM) average causal effects that concept i imparts on FCM concepts
processes with the adjacency matrix of Table 3 are depicted in including itself. Similarly, the average of total effects over column j
Figs. 3 and 4. For illustration, the last column of total effects matrix of FCM normalized total effect matrix can be interpreted as a degree
TM is considered which represents the total effects of HSEE factors of causality by which concept j is influenced. It is noted that
on rate of injury (SP3). because experts are asked to express their opinion about the paired
In both Figs. 3 and 4, 30 data series are depicted each corre- relationships of concepts, the final results of FCM analysis is
sponding to a concept in Table 1. As seen in Fig. 3, the total effects of interpreted as the causal relationship between concepts, as well.
HSEE factors do not show a stable process when M increases. How- Normally, experimental methods are employed to test hypotheses
ever, a steady state is reached in the normalized processes of Fig. 4 concerning cause-and-effect relationship. However, the data
where the values of the normalized total effects remain unchanged needed for experimental analysis should be panel data collected for
for Ms greater than 15. Similar steady state analysis is performed for different companies or for a company in different periods of time.
all concepts in Table 1 and the results are presented in Table 4. Usually, accessing the accurate data due to the lack of data man-
Numbers in Table 4 are the steady state effects of concepts on agement systems especially in developing countries is limited and
each other. One important feature that could be extracted is the this fact restricts the successful application of experimental
average of absolute total effects on or total influences by the other methods. Table 5 shows the averages over rows and columns of the
concepts in FCM. These can be calculated by the means over rows FCM normalized total effect matrix.
and columns of Table 4. For each concept i, the mean of total effects Based on the causality features extracted in Table 5, a diagram of
over row i of FCM normalized total effect matrix (Table 4) is the causality levels (Fig. 5) is developed, which clusters HSEE factors in

Level 1 Level 2 Level 3 Level 4 Level 5 Level 6

H1
H3
S1
H2
S2
S3 MER6
EN1 EN4 ER1 H4
MER4 EN3 MER2 Productivity
MER7 MER3
EN2 MER5
ER6 S5
ER2 S4 Satisfaction
MER1 EN5
ER3 Injuries

ER4
ER5

Fig. 5. The causal levels of HSEE factors and system performance indicators.
1024 S.M. Asadzadeh et al. / Journal of Loss Prevention in the Process Industries 26 (2013) 1015e1026

1.00 map will not be too sensitive to the values chosen for clusters
Overall job satisfaction Workers productivity Injury rate
0.80 thresholds.
The results of causal leveling in Fig. 5 shows that lack of docu-
0.60
mented instructions about works (MER4) can be considered among
0.40 the root causes of low satisfaction, high injuries, and low produc-
0.20 tivity. Other root causes include lack of fresh air (H2), dangerous
Relative effect

and unsuitable work space design (S3), warm places (EN3), ergo-
0.00
nomically unsuitable machines and tools (ER6), and Lack of in-
H2

S4
S5
H1

H3
H4
S1
S2
S3

ER1
ER2
ER3
ER4
ER5
ER6

MER3
MER4
MER5
MER6
MER7
EN1
EN2
EN3
EN4
EN5

MER1
MER2
-0.20 structions and education about safety and accident prevention
-0.40 (MER1). Two HSEE factors of stress and mental pressures (H4) and
work pressures (MER3) are most influenced by other factors and
-0.60
along with low satisfaction and high injury rate contribute to low
-0.80
accounts of workers productivity. Productivity belongs to the last
-1.00 level of causality indicating that it is most influenced by the other
-1.20
factors.
HSEE Factors
Another interesting conclusion that can be drawn from Table 4 is
Fig. 6. The normalized effects of HSEE factors on system performance indicators. to determine the most important factors with regard to system
performance indicators. Fig. 6 shows the total effect of HSEE factors
on system performance indicators. These relative effects provide an
different levels of causality. There are six levels in this diagram. indication on changes occurred in injury rate by the variation in
Factors in each level impart some causality on the factors in their each of HSEE factors and are served as a basis to construct leading
subsequent level. As mentioned before, the average over columns in indicators of system safety performance. Specifically, a leading in-
Table 5 represents the degree by which each concept is under dicator of injury rate can be constructed as in Equation (4).

Leading Inj: Rate ¼ 0:29*H1 þ 0:45*H2 þ 0:28*H3 þ 0:32*H4 þ 0:16*S1 þ 0:31*S2 þ 0:50*S3 þ 0:08S4 þ 0:13*S5 þ 0:25*EN1
þ 0:32EN2 þ 0:47*EN3 þ 0:45*EN4 þ 0:23*EN5 þ 0:29*ER1 þ 0:26*ER2 þ 0:35*ER3 þ 0:33*ER4 þ 0:14*ER5
þ 0:46*ER6 þ 0:65*MER1 þ 0:47*MER2 þ 0:32*MER3 þ 0:80*MER4 þ 0:60*MER5 þ 0:80*MER6
þ 0:35*MER7
(4)

influence of the other concepts in HSEE system FCM. The less this As seen in Equation (4) and Fig. 6, macro-ergonomics factors
degree of influence, the lower the level of causality to which concept i such as lack of instructions and education about safety and accident
belongs. prevention (MER1), lack of documented instructions about works
In the clustering of Fig. 5, the concepts with column mean be- (MER4), unfamiliarity with organization’s rules (MER5), and lack of
tween zero and 0.05 belong to the first level. Level 2 contains proper communications with managers (MER6) are influential
concepts with column mean between 0.05 and 0.15, level 3 with factors that most contribute to the accounts of system performance
column mean between 0.15 and 0.25, level 4 between 0.25 and indicators. It is also noted that the negative signs for S4 and S5 may
0.35, column 5 between 0.35 and 0.45, and finally level 6 contains be interpreted as the effect on injury rate of situation awareness in
concepts with average column greater than 0.45. On the other work place. As the level of awareness of hazards in work place in-
hand, the average of row i in Table 5 represents the degree of effect creases, staff preparedness and readiness increase when encoun-
concept i has on the other concepts in HSEE system FCM. In the tering incidents hence lower rate of injury.
causal leveling of Fig. 5, the level of causality is shown by increased
or decreased indentation. In other words, in each level the con- 5. Qualitative comparison
cepts are organized so that concepts with decreased indentation
have greater row means than the concepts with increased inden- The applicability and usefulness of the proposed approach of
tation indicating more effect on concepts in the subsequent levels. this study is compared with the previous studies in terms of some
These values are intuitively assigned to the clusters thresholds in qualitative features (Table 6). The common characteristic of the
order to form six near-size clusters in the discourse of 0e0.5. studies in Table 6 is that they all focus on the subject of assessment
Moreover, in each cluster indentation is used to differentiate be- and improvement of HSEE risk management systems, most in the
tween the levels of the concepts causality. It is noted that with both process industries. This comparison is performed to highlight the
clustering and indentation in the clusters, the results of causality contribution of the proposed FCM approach. The proposed FCM is

Table 6
The features of the proposed FCM versus other studies.

Reference study Fuzzy modeling Intensive data necessity Integrated HSEE modeling Systemic feedback modeling Causal structure determination
This study O O O O O
Bertolini (2007) O O O
Azadeh et al. (2011) O
Smith and Eloff (2000) O O O
Azadeh et al. (2010) O O
S.M. Asadzadeh et al. / Journal of Loss Prevention in the Process Industries 26 (2013) 1015e1026 1025

capable of dealing fuzzy and subjective assessment in the modeling Azadeh, A., Rouzbahman, M., & Saberi, M. (2010). An artificial intelligent approach
for evaluation of teamwork versus health, safety, environment and ergonomics
environment. This approach does not require intensive data
(HSEE). In G. Salvendy (Ed.), Advances in human factors, ergonomics, and safety in
collection due to its mechanism whereas the artificial intelligent manufacturing and service industries (pp. 1222e1231). USA: CRC Press.
modeling in Azadeh et al. (2011) and Azadeh et al. (2010) need Azadeh, A., Rouzbahman, M., Saberi, M., & Fam, I. M. (2011). An adaptive neural
intensive experimental data. Moreover, this study assesses the network algorithm for assessment and improvement of job satisfaction with
respect to HSE and ergonomics program: the case of a gas refinery, 2011. Journal
influential factors in the integrated HSEE system. The cyclic FCM of of Loss Prevention in the Process Industries, 24(4), 361e370.
this study allows modeling feedbacks and loop effects of the factors. Azadeh, A., Ziaei, B., & Moghaddam, M. (2012). A hybrid fuzzy regression-fuzzy
Finally, the causal structure determination in different levels and cognitive map algorithm for forecasting and optimization of housing market
fluctuations. Expert Systems with Applications, 39, 298e315.
clustering of the factors in different levels of causality is the unique Bertolini, M. (2007). Assessment of human reliability factors: a fuzzy cognitive
feature of this study. maps approach. International Journal of Industrial Ergonomics, 37, 405e413.
Bertolini, M., & Bevilacqua, M. (2010). Fuzzy cognitive maps for human reli-
ability analysis in production systems, in: production engineering and
6. Conclusion management under fuzziness. Studies in Fuzziness and Soft Computing, 252,
381e415.
This study analyzed and assessed the integrated HSEE factors by Bin, W. S., Richardson, S., & Yeow, P. H. P. (2010). An ergonomics study of a semi-
conductors factory in an IDC for improvement in occupational health and safety.
FCM approach. This was achieved through integrating ergonomics and International Journal of Occupational Safety and Ergonomics, 16(3), 345e356.
macro-ergonomics as well as occupational health and safety ar- Burri, G. J., & Helander, M. G. (1991). A field study of productivity improvements in
rangements in an integrated model for assessment of their multi- the manufacturing of circuit boards. International Journal of Industrial Ergo-
nomics, 7, 207e215.
faceted impact on workers’ productivity, injury rate, and workers’
Champoux, D., & Brun, J. J. (2003). Occupational health and safety management in
satisfaction. The direct and indirect effects of HSEE factors on each small size enterprises: an overview of the situation and avenues for interven-
other as well as on system performance indicators were assessed. The tion and research. Safety Science, 41, 301e318.
Changchit, C., & Holsapple, C. W. (2001). Supporting managers’ internal control
result of FCM was used to develop leading indicators useful for pro-
evaluations: an expert system and experimental results. Decision Support Sys-
active management of system performance. As illustrated in this study tems, 30, 437e449.
FCM is a very powerful technique for decision support where there is a Chen, J. R., & Yang, Y. T. (2004). A predictive risk index for safety performance in
need to map knowledge of different experts for system analysis. process industries. Journal of Loss Prevention in the Process Industries, 17, 233e242.
Chytas, P., Glykas, M., & Valiris, G. (2010). Software reliability modelling using fuzzy
In the case study, important health, safety, environment and cognitive maps, In: FUZZY COGNITIVE MAPS. Studies in Fuzziness and Soft
ergonomics factors in gas refineries were determined and then Computing, 247, 217e230.
through the use of cyclic FCM, we mapped the knowledge of ex- Cooper, S. E., Ramey-Smith, A. M., Wreathall, J., Parry, G. W., Bley, D. C., Luckas, W. J.,
et al. (1996). A technique for human error analysis (ATHEANA) (NUREG/CR-6350).
perts about the relationships between HSEE factors and showed the Washington, DC: US Nuclear Regulatory Commission.
cause and effect relations between these factors and then con- Craiger, J., Goodman, D., Wiss, R., & Butler, B. (1996). Modeling organizational
structed leading indicators for system performance. Based on the behavior with fuzzy cognitive maps. International Journal of Computational In-
telligence and Organizations, 1, 120e123.
causality features extracted from FCM, a diagram of causality levels Dickerson, J., & Kosko, B. (1994). Virtual worlds as fuzzy cognitive map. Presence,
was developed, which categorized HSEE factors in different levels 3(2), 173e189.
of causality. Dornyei, Z. (2007). Research methods in applied linguistics. Oxford: Oxford University
Press.
The results of this study showed that macro-ergonomics factors
Dul, J., de Vries, H., Verschoof, S., Eveleens, W., & Feilzer, A. (2004). Combining
such as lack of instructions and education about safety and accident economic and social goals in the design of production systems by using ergo-
prevention, lack of documented instructions about work, unfamil- nomics standards. Computers & Industrial Engineering, 47(2e3), 207e222.
Flanagan, R. J., Strauss, G., & Ulman, L. (1974). Worker discontent and work place
iarity with organization’s rules, and lack of proper communications
behavior. Industrial Relations: A Journal of Economy and Society, 13(2), 101e123.
with managers are important influential factors that most Ghaderi, S. F., Azadeh, A., Pourvalikhan Nokhandan, B., & Fathi, E. (2012). Behavioral
contribute to the accounts of system performance indicators. simulation and optimization of generation companies in electricity markets by
fuzzy cognitive map. Expert Systems with Applications, 39, 4635e4646.
Gyekye, S. A., & Salminen, S. (2010). Organizational safety climate and work experi-
Acknowledgements ence. International Journal of Occupational Safety and Ergonomics, 16(4), 431e443.
Health and Safety Executive. (2013). Understanding ergonomics at work. Retrieved
The authors are grateful for the valuable comments and sug- from http://www.hse.gov.uk/pubns/indg90.pdf.
Hollnagel, E. (1998). Cognitive reliability and error analysis method. Oxford: Elsevier
gestion from the respected reviewers. Their valuable comments Science Ltd.
and suggestions have enhanced the strength and significance of our Kardaras, D., & Karakostas, B. (1999). The use of cognitive maps to simulate the
paper. This study was supported by a grant from University of information systems strategic planning process. Information and Software
Technology, 41, 197e210.
Tehran (Grant No. 8106013/1/12). The authors are grateful for the Kosko, B. (1986). Fuzzy cognitive maps. International Journal of Man-Machine
support provided by the College of Engineering, University of Studies, 24, 65e75.
Tehran, Iran. Kosko, B. (1988). Hidden patterns in combined and adaptive knowledge networks.
International Journal of Approximate Reasoning, 2, 337e393.
Levine, D. I., & Toffel, M. W. (2010). Quality management and job quality: how the
References ISO 9001 standard for quality management systems affects employees and
employers. Management Science, 56(6), 978e996.
Axelord, R. (Ed.). (1976). Structure of decision: The cognitive maps of political elites. Papageorgiou, E., Stylios, C., Groumpos, & Peter. (2003). Fuzzy cognitive map
New Jersey: Princeton Univ. Press. learning based on nonlinear Hebbian rule, AI 2003: ADVANCES IN ARTIFICIAL
Ayoub, M. A. (1990a). Ergonomics deficiencies: I. Pain at work. Journal of Occupa- INTELLIGENCE. Lecture Notes in Computer Science, 2903, 256e268.
tional Medicine, 32(1), 52e57. Ruan, D., & Mkrtchyan, L. (2012). Using Belief degree-distributed fuzzy cognitive
Ayoub, M. A. (1990b). Ergonomics deficiencies: II. Probable causes. Journal of maps for safety culture assessment. In Y. Wang, & T. Li (Eds.), Practical appli-
Occupational Medicine, 32(2), 131e136. cations of intelligent systems, advances in intelligent and soft computing (pp. 501e
Azadeh, A., Fam, I. M., & Azadeh, M. A. (2009). Integrated HSEE management sys- 510). Berlin, Heidelberg: Springer.
tems for industry: a case study in gas refinery. Journal of the Chinese Institute of Saksvik, P., & Nytr, K. (1996). Implementation of internal control (IC) of health, envi-
Engineers, 32(2), 235e241. ronment and safety (HES) in Norwegian enterprises. Safety Science, 23(1), 53e61.
Azadeh, A., Fam, I. M., Khoshnoud, M., & Nikafrouz, M. (2008). Design and imple- Shikdar, A. A., & Sawaqed, N. M. (2003). Worker productivity, and occupational
mentation of a fuzzy expert system for performance assessment of an inte- health and safety issues in selected industries. Computers & Industrial Engi-
grated health, safety, environment (HSE) and ergonomics system: the case of a neering, 45(4), 563e572.
gas refinery. Information Sciences, 178, 4280e4300. Shikdar, A. A., & Sawaqed, M. N. (2004). Ergonomics, occupational health and safety
Azadeh, A., Fam, I. M., Nouri, J., & Azadeh, M. A. (2008). Integrated health, safety, in the oil industry: a managers’ response. Computers & Industrial Engineering, 47,
environment and ergonomics management system (HSEE-MS): an efficient 223e232.
substitution for conventional HSE-MS. Journal of Scientific and Industrial Smith, E., & Eloff, J. (2000). Cognitive fuzzy modeling for enhanced risk assessment
Research, 67, 403e411. in a health care institution. IEEE Intelligent Systems, 15(2), 69e75.
1026 S.M. Asadzadeh et al. / Journal of Loss Prevention in the Process Industries 26 (2013) 1015e1026

Stach, W., Kurgan, L., Pedrycz, W., & Reformat, M. (2005). Genetic learning of fuzzy Toriizuka, T. (2001). Application of performance shaping factors (PSF) for work
cognitive maps. Fuzzy Sets and Systems, 153, 371e401. improvement in industrial plant maintenance tasks. International Journal of
Swain, A. D., & Guttmann, H. E. (1983). Handbook of reliability analysis with emphasis Industrial Ergonomics, 28, 225e236.
on nuclear plant applications. Technical report NUREG/CR-1278. Washington, Wurzelbacher, S., & Jin, Y. (2011). A framework for evaluating OSH program effective-
DC: NUclear REGulatory Commission. ness using leading and trailing metrics. Journal of Safety Research, 42(3), 199e207.
Taber, R. (1991). Knowledge processing with fuzzy cognitive maps. Expert Systems Zio, E. (2009). Reliability engineering: old problems and new challenges. Reliability
with Applications, 2, 83e87. Engineering and System Safety, 94, 125e141.

You might also like