751
1 INTRODUCTION
The International Maritime Organization (IMO), the
International Labour Organization (ILO), the Ship
Classification Societies (IACS), and the
implementation of the 1998 ISM Code 1998, as an
international standard for the safe operation of ships
and the advances in technology, have issued many
rules and standards concerning human errors. After
their implementation, the number of human errors in
maritime accidents was significantly reduced (Akyuz,
Celik, and Cebi 2016; Hetherington, Flin, and Mearns
2006; Lee and Chung 2018; Tzannatos and Kokotos
2009). Nevertheless, despite the continuous
improvement, it is still found that human error
influences maritime accidents (Bowo and Furusho
2019b; Kokotos and Linardatos 2011). According to
the European Maritime Safety Agency (EMSA),
human actions are the most common factor
contributing to maritime accidents, accounting for
approximately 66% of the total of 4104 accidents
analyzed (EMSA 2019). Moreover, it is also supported
by other studies that the percentage of human error
involved in maritime accidents is 80% (Graziano,
Teixeira, and Guedes Soares 2016; Soares and Teixeira
2001; Sotiralis et al. 2016).
A Modified HEART – 4M Method with TOPSIS for
A
nalyzing Indonesia Collision Accidents
L.P. Bowo
, R.E. Prilana & M. Furusho
Kobe University, Kobe, Japan
ABSTRACT: Human error is recognized as the most common factor that causes maritime accidents. The human
error assessment and reduction technique (HEART) is a human reliability assessment (HRA) that has been
widely applied in various industries. Furthermore, the HEART – 4M method has been proposed to assess
maritime accidents. The HEART – 4M method can clearly define the relationship between man, machine,
media, and management factors and the human error. However, the calculation process to determine the
weight of every selected error-producing condition (EPC) suffers from the uncertainty of the assessor's
estimation in practical applications, which may affect the objectivity of its result. In this study, a modification of
the HEART – 4M method with the technique for order preference by similarity to ideal solution (TOPSIS) is
proposed. TOPSIS is a multi-criteria decision making (MCDM) tool. This study aims to develop the HEART –
4M method to make it more comprehensive and objective when assessing maritime accidents. First, the
parameter of the generic task is determined as in the conventional HEART method. Second, the causal factors
are converted to the suitable EPC – 4M, and there are four classification factors for the 38 standard EPCs, which
are divided into man, machine, media, and management factors. Third, the TOPSIS is applied to handle the
problems of interdependencies and interaction among EPC – 4M and the uncertainty that exists in the
assessor´s judgment. The proportion effect of each EPC – 4M is determined through TOPSIS by considering the
correlation among EPC – 4M. Finally, thirteen collision data obtained from the National Transportation and
Safety Committee of Indonesia are assessed to apply the proposed method.
http://www.transnav.eu
the
International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 14
Number
3
September 2020
DOI:
10.12716/1001.14.03.30
752
Besides, the human error is recognized as the
predominant cause not only in maritime accidents but
also in many other domains, such as railway
transportation (Gibson et al. 2013; Wang, Liu, and Qin
2018a), nuclear power plants (Park, Arigi, and Kim
2019), aviation (Kirwan and Gibson 2009), and
healthcare services (Castiglia, Giardina, and
Tomarchio 2015). Thus, numerous researchers and
practitioners have developed alternative models and
theories related to the human reliability analysis
(HRA) (Akyuz et al. 2016; Bowo, Mutmainnah, and
Furusho 2017; Dsouza and Lu 2017; Wang et al.
2018a). The HRA has three purposes: identification of
human errors, prediction of future risk probability,
and reduction of this probability (Kirwan 1996). The
development of HRA comprises three different
generations (Wang, Liu, and Qin 2018b). In the first
generation, in the 1980s, the HRA was developed to
predict and calculate the probability of human error,
and it focused on the skill and rule base level of
human action. The first generation included the
following methodologies: technique for human error
rate prediction (THERP), accident sequence
evaluation program (ASEP), human error assessment
and reduction technique (HEART), and simplified
plant analysis risk Human reliability assessment
(SPAR-H). The second-generation methodologies
considered the influence of internal and external
contexts on the error and the cognitive context that
may influence the system operation. A technique for
human event analysis (ATHEANA) and the cognitive
reliability and error analysis method (CREAM) were
included in the second generation. Finally, the third
generation, which includes the present method and
the developments from previous generations, aims at
being more suitable for particular industries.
HEART is a simple, flexible, and effective method
for determining the human error involved in
accidents. Therefore, it has been used in various
industries with complex systems, such as nuclear
power plants, railway transportation, aviation, off-
shore platforms, and the maritime industry (Akyuz et
al. 2016; Bowo and Furusho 2019a; Castiglia et al.
2015; Deacon et al. 2013; Gibson et al. 2013; Wang et
al. 2018b). There have been some developments of the
HEART method to handle its limitations, especially
for calculating the value of human error probability
(HEP). The fault tree analysis and fuzzy set theory
were hybridized with the HEART method to
determine the HEP in irradiation plants (Casamirra et
al. 2009; Castiglia and Giardina 2011). The fuzzy set
theory was also employed to assess the HEP in
hydrogen refueling stations (Castiglia and Giardina
2013). In the maritime industry, the HEART method
has been integrated with the analytic hierarchy
process (AHP) method to determine the specific value
of an error producing condition (EPC) (Akyuz and
Celik 2015). In the railway industry, a combination of
the fuzzy analytic network process (FANP) and
HEART method is utilized to determine the weight of
the assessed proportion effect (APE) for HEP
calculation (Wang et al. 2018b). The fuzzy logic theory
has been combined with the HEART method to solve
the linguistic expressions of expert elicitations to
determine the appropriate weight to an EPC
( Kumaret al., 2017).
In light of the above explanation, many
developments of the HEART method have been
realized in various industries. Although several
developments of the HEART method have overcome
its limitations and shortcomings, most of these
developments lack consideration of the relation
between EPCs. In the maritime working environment,
machinery, environment, and management can also
influence the human condition to make judgments
and control the situation. Furthermore, these factors
have a strong relationship with human factors. This
condition has been described in the HEART – 4M
method, where the EPCs are categorized into four
factors: man, machine, media, and management.
However, the relationship between the factors and the
HEP calculation process is still an issue. To remedy
this gap, this study proposes a HEART – 4M method
by combining it with the technique for order
preference by similarity to ideal solution (TOPSIS) to
evaluate the HEP in maritime accidents. The TOPSIS
is introduced to handle the determination of the APE
and the relation between factors. This paper presents
a modified HEART – 4M method combined with
TOPSIS to assess human error probability for collision
accidents in Indonesia.
2 METHODOLOGY
This study proposes a modified method to evaluate
HEP by incorporating the HEART method, 4M
framework, and TOPSIS method in maritime collision
accidents. Therefore, a description of these methods is
provided below.
2.1 HEART method
HEART, established by Williams (1988), is a robust
method to evaluate the HEP with defined error
probability values. There are two fundamental
parameters described in the HEART method: the
generic task (GT) and the EPC. The GT parameter
consists of nine qualitative descriptions of the
appropriate task in the accident process, which is
carried out by the assessor when analyzing the case.
The GT also provides values of generic error
probability, named nominal human unreliability
(NHU), for every GT. The second fundamental
parameter is the EPC, which indicates the relevant
performance shaping factors for humans during the
course of a task and can affect the value of HEP. The
EPC can be any internal human feature or be related
to other factors such as machine, management, and
environment. There are 38 EPCs defined in the
HEART method, and every EPC is provided with a
multiplied number, which will later be used to
calculate the HEP.
In light of the above, the calculation formula to determine
the value of EPC is shown below:
( )
11
value i i
i
HEP NHU EPC APE

=× −+


∏
(1)
753
Table 1. Generic Tasks (GT).
__________________________________________________________________________________________________
Generic Tasks (GT)
Code Type of work Condition NHU
__________________________________________________________________________________________________
A Totally unfamiliar Performing the work at speed with no real idea of likely 0.55
consequences.
B Restore the system to an original state on a Doing work without supervision or procedures. 0.26
single attempt
C Complex task It requires a high level of comprehension and skill. 0.16
D A fairly simple task Performing the work rapidly or given scant attention. 0.09
E The routine, highly practiced, rapid task Involving a relatively low level of skill. 0.02
F Restore a system to original An error occurred even though following procedures 0.003
with some verification.
G Entirely familiar, highly practiced, routine task Without the benefit of significant job aids. 0.0004
occurring several times per hour, performed to
the highest possible standards by a highly
motivated, highly trained, and experienced person,
totally aware of implications of failure, with time
to correct the potential error
H Respond correctly to the system command Even when there is an augmented or automated 0.00002
supervisory system providing an accurate interpretation
of the system stage.
M Miscellaneous task for which no description can be found. 0.03
__________________________________________________________________________________________________
where NHU is the error probability value of the
relevant GT, and EPCi is the ith (i = 1,2,3, ⋯n) error
producing condition, and the assessed proportion
effect (APE) is a weight that corresponds to the
importance of every EPC. As the EPC influence in the
case becomes more critical, the value of the APE will
be higher.
2.2 HEART – 4M method
The HEART – 4M method is a methodological
extension of the conventional HEART method, which
was introduced by Bowo and Furusho (Bowo and
Furusho 2019b).
The HEART – 4M method is similar to the
conventional HEART method, which consists of
qualitative and quantitative approaches. In the
qualitative approach, the selection of the relevant GT
and NHU for the particular conditions before the
accident uses the same GT parameter as the
conventional HEART. Table 1 presents the GT and
NHU used in this study.
However, in the HEART – 4M method, there is a
categorization of the EPCs into the 4M framework,
which consist of man, machine, media, and
management. In the maritime working condition, the
human condition can be influenced by machine,
media, and management factors while performing
tasks. Moreover, the 38 EPCs that were established by
William describe the working condition, not only the
error exclusively due to the human himself, but also
to the interaction between humans, human–machine
interactions, and working environment conditions.
Table 2 presents the EPC – 4M categorization and the
multiplication number. Therefore, with this
categorization, the relationships between factors and
the involvement of other factors in maritime accidents
are now well addressed. Table 2 lists the
categorization of the EPC and 4M factors and the
multiplication of every EPC that will be used in the
HEP calculation. There are five and four sub-factors
in the man and management factors, respectively.
Furthermore, the quantitative approach to calculate
the result of HEP is based on Formula (1).
As mentioned above, the conventional HEART
and HEART – 4M methods have limitations in
describing the dependencies among EPCs and
determining the weight of the APE to eliminate the
uncertainties in error probability calculation.
Therefore, TOPSIS is applied to modify the HEART –
4M method for developing an assessment to
determine the weight value of the APE.
Table 2. EPC – 4M categorization and the multiplication
_______________________________________________
Man Factors ×
_______________________________________________
Physical limitations
EPC 27 Physical capabilities 1.4
EPC 36 Task pacing 1.06
EPC 38 Age 1.02
Psychological limitations
EPC21 Dangerous incentives 2
EPC28 Low meaning 1.4
EPC 29 Emotional stress 1.3
EPC 31 Low morale 1.2
EPC 34 Low mental workload 1.1
Experience
EPC 1 Unfamiliarity 17
EPC 12 Misperception of risk 4
EPC 22 Lack of experience 1.8
Skill and Knowledge
EPC 7 Irreversibility 8
EPC 9 Technique unlearning 6
EPC 11 Performance ambiguity 5
EPC 15 Operator inexperience 3
EPC 20 Educational mismatch 2
Health
EPC 30 Ill-health 1.2
EPC 35 Sleep cycles disruption 1.1
_______________________________________________
Machine Factors
_______________________________________________
EPC 3 Low signal-noise ratio 10
EPC 8 Channel overload 6
EPC 23 Unreliable instruments 1.6
_______________________________________________
Media Factors
_______________________________________________
EPC 33 Poor environment 1.15
_______________________________________________
Management Factors
_______________________________________________
754
Communication
EPC 13 Poor feedback 4
EPC 14 Delayed/incomplete feedback 3
EPC 16 Impoverished information 3
EPC 18 Objectives conflict 2.5
EPC 19 No diversity of information 2.5
Coordination
EPC 2 Time shortage 11
EPC 6 Model mismatch 8
EPC 10 Knowledge transfer 5.5
EPC 24 Absolute judgments required 1.6
EPC 25 Unclear allocation of function 1.6
EPC 37 Supernumeraries/ lack of human resources 1.03
Monitoring
EPC 17 Inadequate Checking 3
EPC 26 Progress tracking lack 1.4
Procedures
EPC 4 Features over-ride allowed 9
EPC 5 Spatial and functional incompatibility 8
EPC 32 Inconsistency of displays 1.2
_______________________________________________
2.3 TOPSIS
TOPSIS is a multi-criteria decision-making tool.
TOPSIS was introduced in 1981 by Hwang and Yoon
(1981), and it has been widely used in complex
decision-making problems in various domains.
TOPSIS aims to calculate the importance weight of
alternatives through their similarity with an ideal
solution (Krohling and Pacheco 2015; Olson 2004).
TOPSIS comprises the set of processes described
below. The first process constructs a pair-wise
comparison matrix. The Saaty's 1–9 linguistic relative
importance scale is used (Saaty 1985).
Table 3. Saaty's pair-wise comparison scale.
_______________________________________________
Importance scale Definition
_______________________________________________
1 Equal importance
3 Moderate importance
5 Strong importance
7 Extreme importance
9 Absolute extreme importance
2, 4, 6, 8 Intermediate values
_______________________________________________
1 A pair-wise comparison matrix (D) can be
established in accordance with Formula (2). In the
formula, xij (i = 1, 2, …, m, j = 1, 2, …, n) has the
relative importance of the ith element compared to
the jth. In this study, every selected EPC will be
compared to the other selected EPCs to determine
the interdependencies of EPCs. By comparing
these EPCs, it can be observed that every EPC is
related to each other, and there will be a tendency
for an EPC to be a major factor in an accident.
(2)
1, 1 / , 0
ii ij ji ji
x x xx= = ≠
2 The normalized decision matrix is constructed and
weighted.
Normalized decision matrix
To construct the normalized decision matrix, first, the
attribute weight (w
i) for each EPCi must be obtained
by utilizing Formula (3).
2
1
m
i ij
i
wx
=
=
∑
(3)
After obtaining the attribute weight, the
normalized decision matrix (r_ij) is constructed by
dividing the value from the pair-wise comparison
matrix to the attribute weight, as shown in Formula
(4).
ij
ij
i
x
r
w
=
(4)
Weighted normalized decision matrix.
ij ij ij
p rx= ×
(5)
3 The ideal and negative ideal solutions are
determined.
Ideal solution
2
( )
ij ij i max
d pp
+
= −
(6)
Negative ideal solution
2
( )
ij ij i min
d pp
−
= −
(7)
4 The separation from the ideal solution is
determined.
2
1
()
n
i ij
j
dd
++
=
=
∑
(8)
5 The separation from the negative ideal solution is
determined.
2
1
()
n
i ij
j
dd
−−
=
=
∑
(9)
6 Relative closeness to the ideal solution.
i
i
ii
d
dd
ξ
−
+−
=
+
(10)
7 Normalization.
The summation of all the EPC ideal solution
values is not one, it is often more than one and
sometimes even less than 1. Thus, it needs to be
normalized before using this value for the HEP
calculation. The last value used in the HEP calculation
is the normalization value (N) to be the weight in the
APE. This value shows which EPC has the highest
755
value of weight, which implicates this is the main
factor of the accident because its particular EPC is the
most important compared with other EPCs. If the
weight is approved, then it can be used for the HEP
calculation. Therefore, in this study, the highest value
of EPC was named the top of the EPC series. Formula
(11) shows the calculation formula for the
normalization value.
N
i
ξ
=
∑
(11)
8 Consistency verification
The next step proves the consistency of data. This
step verifies whether the comparison pair-wise matrix
is consistent or not. The consistency index (CI) can be
calculated using the following formula:
1
n
ij max i
j
xN N
λ
=
=
∑
(12)
1
max
n
CI
n
λ
−
=
−
(13)
A consistency verification calculation is needed to
specify a reasonable consistency. The consistency
ratio (CR) value was ≤ 0.10. Otherwise, the expert
judges will be revised to obtain consistent results.
CI
CR
RI
=
(14)
In the equation, RI stands for random index. It is
subjected to a number of items that are compared in
the matrix. The RI values are provided in Table 4.
Table 4. Random index values (Saaty 1994).
_______________________________________________
n 1 2 3 4 5
_______________________________________________
RI 0 0 0.58 0.90 1.12
_______________________________________________
n 6 7 8 9 10
_______________________________________________
RI 1.24 1.32 1.41 1.45 1.49
_______________________________________________
3 RESULTS
In this study, data on maritime collision accidents
from the Indonesian National Transportation and
Safety Committee in the period of 2009–2018 were
used. In total, 13 data sets were collected, and 23 ships
were involved in the analysis. The types of ships
involved in collision accidents were container ships,
bulk carriers, tanker ships, cargo ships, passenger
ships, and tug boats. The cases that have been
analyzed are ships with more than 500 GT.
3.1 Generic Task
From the 23 ships involved in collision accidents in
Indonesia, Table 5 presents the tabulation of the
selected GT. The most common situations
encountered by Indonesian ships are routine, highly
practiced, and rapid tasks that involve a relatively
low level of skill. This shows that there are 17 ships
that had the same working situation before the
accidents occurred.
All of the collision accidents occurred when the
ship sailed in or out to the destination port, which has
a high density and congested traffic. Nineteen cases
started as a fairly simple task, under the condition
where the navigation team received help from the tug
boat or pilot to enter or exit the port. The condition
considers that the pilot and tug boat pilots are already
familiar with the water's situation. It is included in the
category of a fairly simple task, but it was performed
rapidly and received scant attention. In addition,
there were four ships that entered and exited the port
without the tug boat or local pilot assistance, although
there are rules that govern this. This type of situation
requires a high level of skill and is included in the
complex task type because assistance is required to
carry out this job properly.
Table 5. Generic task result
_______________________________________________
Code Type of work Total
_______________________________________________
D A fairly simple task 19
C Complex task 4
_______________________________________________
3.2 EPC – 4M
There are 101 selected EPCs for the 23 ships that have
been assessed. The total of EPCs in the man factor is
47, which are divided into four sub-factors: physical,
psychological, experience, and skill and knowledge.
Misperception of risk is the most common EPC found
in Indonesian cases. Moreover, the educational
mismatch was also found, and four cases were
identified. In these cases, the seafarer did not have
qualified education to work onboard. However, due
to the shortage of crew, unqualified seafarers were
employed.
Management factors have more EPCs than the
man factors. There were 54 EPCs found, consist of
communication, coordination, and monitoring sub-
factors that affect collision accidents in Indonesia. The
most common EPCs found belong to monitoring sub-
factors in management factors. They correspond to
EPC 17, inadequate verification for 14 ships, and EPC
26, lack of progress tracking for 12 ships.
Moreover, five ships had machine factors due to
unreliable instruments. Only one case has media
factors. The details of EPCs found in Indonesia
collision accident cases are summarized in Table 6.
756
Table 6. EPC – 4M results
_______________________________________________
Man Factors
_______________________________________________
Physical
EPC 36 Task pacing 2
Psychological
EPC 21 Dangerous incentives 4
EPC 28 Low meaning 2
EPC 29 Emotional stress 2
EPC 34 Low mental workload 4
Experience
EPC 1 Unfamiliarity 1
EPC 12 Misperception of risk 9
EPC 22 Lack of experience 6
Skill and Knowledge
EPC 20 Educational mismatch 4
EPC 9 Technique unlearning 1
EPC 11 Performance ambiguity 6
_______________________________________________
Machine Factors
_______________________________________________
EPC 23 Unreliable instruments 5
_______________________________________________
Management Factors
_______________________________________________
Communication
EPC 10 Knowledge transfer 4
EPC 13 Poor feedback 9
EPC 16 Impoverished information 3
EPC 18 Objectives conflict 2
EPC 19 No diversity of information 1
Coordination
EPC 2 Time shortage 4
EPC 24 Absolute judgments required 1
EPC 37 Supernumeraries/lack of human resources 4
Monitoring
EPC 17 Inadequate verification 14
EPC 26 Lack of progress tracking 12
_______________________________________________
Media Factors
_______________________________________________
EPC 33 Poor environment 1
_______________________________________________
Total 101
_______________________________________________
3.3 HEP Calculation
In this section, we consider one of the cases to be the
calculation example of this proposed method. The
following calculation description is from case number
one, with the following details: this accident occurred
on May 22nd, 2009, at 17:28 in Madura Strait,
Surabaya. The weather conditions at that time were
fine, with calm winds and currents of 1.8 knots from
the west. This accident involved two ships, a
container ship of 5,283 GT and a general cargo ship of
8,639 GT. However, the accident report on NTSC only
stated the container ship condition. Therefore, the
analysis of case number one only assessed one ship.
In case one, there are five EPCs selected, which
comprised EPC 11, EPC 21, EPC 12, EPC 29, and EPC
1. To determine the APE weight of each of these EPCs,
the data are processed using TOPSIS, as follows:
1 Pair-wise comparison matrix (D)
After selecting the EPCs that caused the accident
in the accident report, the next step in calculating the
APE weight value is constructing the pair-wise
comparison matrix, as presented in Table 7. In the
matrix, every EPC is selected by putting the
importance scale and using Formula (2) to calculate
the proportion.
The attribute weight (w
i) in this table is calculated
using Formula (3). The attribute weight value is used
in the next step to construct the normalized decision
matrix.
Table 7. Pair-wise comparison matrix and attribute weights
(w
i)
_______________________________________________
EPC11 EPC21 EPC12 EPC29 EPC1 wi
_______________________________________________
EPC11 1 0.33 3 3 0.5 4.40
EPC21 3 1 0.2 0.33 0.25 3.20
EPC12 0.33 5 1 0.2 0.33 5.12
EPC29 0.33 3 5 1 0.25 5.93
EPC1 2 4 3 4 1 6.78
_______________________________________________
2 The normalized decision matrix is constructed and
weighted.
Normalized decision matrix
After calculating the attribute weight (w
i), then the
normalized decision matrix is constructed, Table 8, by
utilizing Formula (4).
Table 8. Normalized decision matrix
_______________________________________________
EPC11 EPC21 EPC12 EPC29 EPC1
_______________________________________________
EPC11 0.23 0.08 0.68 0.68 0.11
EPC21 0.94 0.31 0.06 0.10 0.08
EPC12 0.07 0.98 0.20 0.04 0.07
EPC29 0.06 0.51 0.84 0.17 0.04
EPC1 0.29 0.59 0.44 0.59 0.15
_______________________________________________
Weighted normalized decision matrix.
In the weighted normalized decision matrix, in
Table 9, the maximum weight (p
(i max)) and the
minimum weight (p
(i min)) for every EPC, listed in Table
10, are used. The maximum weight is used to
calculate the ideal solution matrix, and the minimum
weight will be used for the negative-ideal solution
matrix.
Table 9. Weighted normalized decision matrix
_______________________________________________
EPC11 EPC21 EPC12 EPC29 EPC1
_______________________________________________
EPC11 0.23 0.03 2.05 2.05 0.06
EPC21 2.82 0.31 0.01 0.03 0.02
EPC12 0.02 4.88 0.20 0.01 0.02
EPC29 0.02 1.52 4.22 0.17 0.01
EPC1 0.59 2.36 1.33 2.36 0.15
_______________________________________________
Table 10. Maximum and minimum weight
_______________________________________________
MAX MIN
_______________________________________________
2.05 0.03
2.82 0.01
4.88 0.01
4.22 0.01
2.36 0.59
_______________________________________________
3 The ideal and negative ideal solutions are
determined.
Ideal solution matrix and separation from the ideal
solution d
i
+
The ideal solution is the maximum limit that can
be reached for every EPC from the calculation, as
presented in Table 11.
757
Table 11. Ideal solution matrix and separation from the ideal
solution d
i
+
.
_______________________________________________
EPC11 EPC21 EPC12 EPC29 EPC1
_______________________________________________
EPC11 3.31 4.08 0 0 3.95
EPC21 0.00 6.27 7.86 7.74 7.82
EPC12 23.59 0.00 21.93 23.72 23.59
EPC29 17.61 7.28 0.00 16.38 17.68
EPC1 3.13 0 1.06 0 4.89
d
i
+
23.82 8.81 15.43 23.92 28.97
_______________________________________________
Negative ideal solution matrix and separation
from negative ideal solution d
i
-
.
The negative ideal solution is the minimum value
that can be reached for every EPC from the
calculation, as presented in Table 12.
Table 12. Negative ideal solution matrix and separation
from the negative ideal solution d
i
-
.
_______________________________________________
EPC11 EPC21 EPC12 EPC29 EPC1
_______________________________________________
EPC11 0.04 0 4.08 4.08 0.0009
EPC21 7.86 0.09 0.00 0.00 0.00
EPC12 0.00 23.72 0.04 0.00 0.00
EPC29 0.00 2.27 17.68 0.02 0.00
EPC1 0.00 3.13 0.54 3.13 0.20
d_i^- 3.95 14.61 11.17 3.62 0.10
_______________________________________________
4 Relative closeness to the ideal solution and
normalization
After obtaining the result of the ideal and negative
ideal solution, the relative closeness to the ideal
solution must be calculated using Formula (10).
Because the summation of all the values of relative
closeness to the ideal solution is more than 1 in this
example, it needs to be normalized to the total weight.
Table 13 lists the values of relative closeness to the
ideal solution and its normalization value.
Table 13. Relative closeness to ideal solution and
normalization
_______________________________________________
EPC11 EPC21 EPC12 EPC29 EPC1 Total
_______________________________________________
i
ξ
0.14 0.62 0.42 0.13 0.0034 1.32
N 0.11 0.47 0.32 0.10 0.0026 1.00
_______________________________________________
5 Consistency verification
Before using the normalization value in the HEP
calculation, the consistency of the value given in the
pair-wise comparison matrix needs to be verified. The
CI can be calculated using Formula (12), as presented
in Table 14. The RI value was established by Saaty
because, in this case, the number of EPCs found was
five, and the RI assigned for calculating the CR was
1.1086. If CR ≤ 0.1, the normalization can be
accepted and used in the HEP calculation.
Table 14. Consistency check
_______________________________________________
CI RI CR
_______________________________________________
0.07 1.1086 0.061
_______________________________________________
6 HEP Calculation
Table 15 presents the calculation example of the
HEP result for case number one. The GT that was
selected for the condition before the accident is a
complex task that requires a high level of
comprehension and skill, which has an NHU of 0.16
because it enters the Madura strait without guidance
from a local tug boat. Table 15 presents the EPC series
in case one, which has EPC 21 as the top of EPC
series, followed by EPC 12, EPC 11, EPC 29, and EPC
1.
Table 15. HEP Calculation
_______________________________________________
TOP BODY
_______________________________________________
EPC 21 EPC 12 EPC 11 EPC 29 EPC 1
_______________________________________________
× APE × APE × APE × APE × APE
2 0.47 4 0.32 5 0.11 1.3 0.09 17 0.003
_______________________________________________
HEP Value 0.71
_______________________________________________
Figure 1 shows the results of the twenty-three
ships that were assessed using the proposed methods,
HEART – 4M and TOPSIS. The figure shows that one
or two ships are assessed in one case. The value of
HEP varied for each case. The average value of the
Indonesian collision accident in Indonesia was 41%.
The value of HEP can vary owing to the differences in
the selected GT and the number of EPCs found in
each case. If the selected GT has a higher NHU, the
value of HEP can also be higher.
Figure 1. HEP Value of Indonesian collision accidents and
average of the HEP value
4 DISCUSSION
Human factors are still the main factors of collision
accidents in Indonesia. The analysis of the reviewed
collision accidents in Indonesia shows that most
accidents occurred during fairly simple tasks, which
were rapidly performed and received scant attention.
It means that the seafarers did not pay sufficient
attention during their onboard work to manage
satisfactory watchkeeping tasks. They thought they
were familiar with the situation, and thus they tended
to underestimate the task. Based on the EPC – 4M
result, the management factors influenced the
condition of humans during their work. The
management factor becomes the most common EPC –
4M factor found in these cases, besides the man factor
itself. This means that the tasks related to
management and which require good teamwork, such
758
as monitoring, communication, and coordination,
must receive more attention. However, the EPCs
belonging to man factors are also of concern, as it has
been found that many of these factors influence the
accidents. The Indonesian seafarer must be trained
and educated well before working on board, and all
the stakeholders of the Indonesian shipping
companies have to obey the rules that have been
issued by the authority for the safety at sea. In some
cases, it was found that some seafarers did not have
sufficient qualifications to work onboard, yet they
worked, and did not have enough capacity to handle
a certain condition to prevent accidents.
The HEART method is a robust tool for analyzing
the human error probability. However, this method
has some limitations to connect each EPC that has an
attachment to other factors and to calculate the HEP
value in the maritime industry. To overcome these
limitations, first, the HEART method has been
combined with the 4M factors to categorize the EPC
into man, machine, media, and management factors
(Bowo, Prilana, and Furusho 2019). This
categorization can define all the 38 EPCs established
by William in 1986 into the 4M factors, which are
related to the maritime industry's working
environment. This 4M factors are related to each other
because each factor can also influence other factors.
Second, TOPSIS is used to determine the weight of the
APE for every selected EPC in the case by considering
the relation of every EPC.
Finally, a hybrid method that integrates HEART –
4M and TOPSIS to calculate the maritime accidents in
Indonesia was proposed. The integration of these
methods suggests the relation between the EPC and
the 4M method along with the dependencies among
them. The problem with the relationships between
factors and the involvement of other factors in
maritime accidents is now well addressed. The
TOPSIS method also helps the assessor to determine
the weight of the APE for every selected EPC.
5 CONCLUSION
HRA is considered as a tool to determine the
probability of human error and help the decision-
maker to develop a mitigation process to avoid the
same situation in the future. The purpose of this
paper is to introduce a new method for quantifying
the HEP in maritime accidents, in this case, collision
accidents. Owing to some limitations of the HEART
method, a number of developments of this method
have been conducted. In this study, the HEART – 4M
method, based on the TOPSIS method, is proposed to
overcome the limitation of the HEART method for
analyzing maritime accident cases. The TOPSIS
method can be used to obtain the uncertainty of
weight for every EPC and determine the
dependencies among EPCs to determine the most
influential EPC in a particular maritime accident.
Furthermore, the result of the analysis of Indonesian
maritime collision accidents shows that the most
common GT is a fairly simple task that is rapidly
performed and receives scant attention. Further, the
EPCs of management factors are the most common
causal factors found in these accidents. In conclusion,
the hybrid method proposed in this study provides a
practical tool to determine the value of HEP in
maritime accidents.
REFERENCES
Akyuz, Emre and Metin Celik. 2015. “A Methodological
Extension to Human Reliability Analysis for Cargo Tank
Cleaning Operation on Board Chemical Tanker Ships.”
Safety Science 75:146–55.
Akyuz, Emre, Metin Celik, and Selcuk Cebi. 2016. “A Phase
of Comprehensive Research to Determine Marine-
Specific EPC Values in Human Error Assessment and
Reduction Technique.” Safety Science 87:63–75.
Bowo, Ludfi Pratiwi and Masao Furusho. 2019a. “Usability
of Human Error Assessment and Reduction Technique
with a 4M Framework (HEART-4M) - A Case Study on
Ship Grounding Accidents.” Journal of ETA Maritime
Science 7(4):0.
Bowo, Ludfi Pratiwi and Masao Furusho. 2019b. “Usability
of Human Error Assessment and Reduction Technique
with a 4M Framework (HEART–4M) – A Case Study on
Ship Grounding Accidents.” Journal of ETA Maritime
Science 7(October):266–79.
Bowo, Ludfi Pratiwi, Wanginingastuti Mutmainnah, and
Masao Furusho. 2017. “The Development of Marine
Accidents Human Reliability Assessment Approach:
HEART Methodology and MOP Model.” TransNav, the
International Journal on Marine Navigation and Safety
of Sea Transportation 11(2):63–68.
Bowo, Ludfi Pratiwi, Ramdhani Eka Prilana, and Masao
Furusho. 2019. “A Hybrid Methodology for Maritime
Accident Analysis: The Case of Ship Collision.”
Casamirra, M., F. Castiglia, M. Giardina, and E. Tomarchio.
2009. “Fuzzy Modelling of HEART Methodology:
Application in Safety Analyses of Accidental Exposure
in Irradiation Plants.” Radiation Effects and Defects in
Solids 164(5–6):291–96.
Castiglia, F. and M. Giardina. 2011. “Fuzzy Risk Analysis of
a Modern γ-Ray Industrial Irradiator.” Health Physics
100(6):622–31.
Castiglia, F. and M. Giardina. 2013. “Analysis of Operator
Human Errors in Hydrogen Refuelling Stations:
Comparison between Human Rate Assessment
Techniques.” International Journal of Hydrogen Energy
38(2):1166–76.
Castiglia, Francesco, Mariarosa Giardina, and Elio
Tomarchio. 2015. “THERP and HEART Integrated
Methodology for Human Error Assessment.” Radiation
Physics and Chemistry 116:262–66.
Deacon, T., P. R. Amyotte, F. I. Khan, and S. Mackinnon.
2013. “A Framework for Human Error Analysis of
Offshore Evacuations.” Safety Science 51(1):319–27.
Dsouza, Ninochka and Lixuan Lu. 2017. “A Literature
Review on Human Reliability Analysis Techniques
Applied for Probabilistic Risk Assessment in the
Nuclear Industry BT - Advances in Human Factors in
Energy: Oil, Gas, Nuclear and Electric Power
Industries.” Pp. 41–54 in, edited by S. M. Cetiner, P.
Fechtelkotter, and M. Legatt. Cham: Springer
International Publishing.
EMSA. 2019. Annual Overview of Marine Casualties and
Incidents 2019.
Gibson, W., A. Mills, S. Smith, and B. Kirwan. 2013. Railway
Action Reliability Assessment, a Railway-Specific
Approach to Human Error Quantification.
Graziano, A., A. P. Teixeira, and C. Guedes Soares. 2016.
“Classification of Human Errors in Grounding and
Collision Accidents Using the TRACEr Taxonomy.”
Safety Science 86:245–57.
759
Hetherington, Catherine, Rhona Flin, and Kathryn Mearns.
2006. “Safety in Shipping: The Human Element.” Journal
of Safety Research 37(4):401–11.
Hwang, Ching-Lai and Kwangsun Yoon. 1981. “Multiple
Attribute Decision Making: Methods and Applications.”
Springer-Verlag Berlin Heidelberg.
Kirwan, B. 1996. “The Validation of Three Human
Reliability Quantification Techniques - THERP, HEART
and JHEDI: Part 1 - Technique Descriptions and
Validation Issues.” Applied Ergonomics 27(6):359–73.
Kirwan, B. and W. Gibson. 2009. “Controller Action
Reliability Assessment ( CARA ) CARA User Manual.”
(August):1–70.
Kokotos, Dimitrios X. and Dionisios S. Linardatos. 2011.
“An Application of Data Mining Tools for the Study of
Shipping Safety in Restricted Waters.” Safety Science
49(2):192–97.
Krohling, Renato A. and André G. C. Pacheco. 2015. “A-
TOPSIS - An Approach Based on TOPSIS for Ranking
Evolutionary Algorithms.” Procedia Computer Science
55(Itqm):308–17.
Lee, Joohee and Hyun Chung. 2018. “A New Methodology
for Accident Analysis with Human and System
Interaction Based on FRAM: Case Studies in Maritime
Domain.” Safety Science 109(May):57–66.
Maniram Kumar, A., Sivaprakasam Rajakarunakaran, and
V. Arumuga Prabhu. 2017. “Application of Fuzzy
HEART and Expert Elicitation for Quantifying Human
Error Probabilities in LPG Refuelling Station.” Journal of
Loss Prevention in the Process Industries 48:186–98.
Olson, D. L. 2004. “Comparison of Weights in TOPSIS
Models.” Mathematical and Computer Modelling 40(7–
8):721–27.
Park, Jooyoung, Awwal Mohammed Arigi, and Jonghyun
Kim. 2019. “Treatment of Human and Organizational
Factors for Multi-Unit HRA: Application of SPAR-H
Method.” Annals of Nuclear Energy 132:656–78.
Saaty, Thomas L. 1985. “Axiomatization of the Analytic
Hierarchy Process.” 91–108.
Saaty, Thomas L. 1994. “How to Make a Decision: The
Analytic Hierarchy Process.” Interfaces 26(4):19–43.
Soares, C. Guedes and A. P. Teixeira. 2001. “Risk
Assessment in Maritime Transportation.” Reliability
Engineering and System Safety 74(3):299–309.
Sotiralis, P., N. P. Ventikos, R. Hamann, P. Golyshev, and A.
P. Teixeira. 2016. “Incorporation of Human Factors into
Ship Collision Risk Models Focusing on Human Centred
Design Aspects.” Reliability Engineering and System
Safety 156:210–27.
Tzannatos, Ernestos and Dimitris Kokotos. 2009. “Analysis
of Accidents in Greek Shipping during the Pre- and
Post-ISM Period.” Marine Policy 33(4):679–84.
Wang, Weizhong, Xinwang Liu, and Yong Qin. 2018a. “A
Modified HEART Method with FANP for Human Error
Assessment in High-Speed Railway Dispatching Tasks.”
International Journal of Industrial Ergonomics
67(September 2017):242–58.
Wang, Weizhong, Xinwang Liu, and Yong Qin. 2018b. “A
Modified HEART Method with FANP for Human Error
Assessment in High-Speed Railway Dispatching Tasks.”
International Journal of Industrial Ergonomics
67(September 2017):242–58.
Williams, J. C. 1988. “A Data-Based Method for Assessing
and Reducing Human Error to Improve Operational
Performance.” Pp. 436–50 in Conference Record for 1988
IEEE Fourth Conference on Human Factors and Power
Plants,.