Reliability Engineering and System Safety 249 (2024) 110201

Contents lists available at ScienceDirect

Reliability Engineering and System Safety

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

Human and organizational factors analysis of collision accidents between

merchant ships and fishing vessels based on HFACS-BN model
Hong Wang a, b, f, Ning Chen a, b, e, Bing Wu b, c, d, *, C. Guedes Soares g
a
School of Transportation and Logistics Engineering, Wuhan University of Technology, Wuhan 430063, China
b
State Key Laboratory of Maritime Technology and Safety, Wuhan University of Technology, Wuhan 430063, China
c
Intelligent Transport Systems Research Centre, Wuhan University of Technology, Wuhan 430063, China
d
National Engineering Research Centre for Water Transport Safety (WTSC), Wuhan University of Technology, Wuhan 430063, China
e
Sanya Science and Education Innovation Park, Wuhan University of Technology, Sanya 572004, China
f
School of Economics and Management, Huaibei Normal University, Huaibei 235000, China
g
Centre for Marine Technology and Ocean Engineering, Instituto Superior Tecnico, Universidade de Lisboa, Portugal

A R T I C L E I N F O A B S T R A C T

Keywords: Collision accidents between merchant ships and fishing vessels have attracted much attention owing to the
Merchant ships and fishing vessels relatively high frequency and serious consequences, at least in China. This paper proposes a hybrid methodology
Collision accidents incorporating the Human Factors Analysis and Classification System and Bayesian networks to investigate the
Human factors analysis and classification
human and organizational factors of collision accidents between merchant ships and fishing vessels. The kernel
system
of this model is to identify human and organizational factors using a modified Human Factors Analysis and
Bayesian networks
Human and organizational factors Classification System framework based on 443 historical collision accidents from 2013 to 2023 in China, to
Development paths transform relationships of five levels in the Human Factors Analysis and Classification System to the graphical
structure of the Bayesian network and to apply Expectation Maximization algorithm for parameter learning to
obtain the parameters of the Bayesian network. 56 relevant human and organizational factors are identified,
including 11 special influencing factors related to collisions between merchant ships and fishing vessels. The
proposed model is validated using two axioms, key influencing factors are identified through sensitivity analysis,
and development paths of collision accidents are derived through the strength of influence analysis. Conse­
quently, the findings of this study provide valuable insights for the Maritime Administration, fisheries supervi­
sory agencies, shipping companies, fishermen, and other stakeholders in formulating effective strategies to
prevent collision accidents between merchant ships and fishing vessels.

1. Introduction River. Yu et al. [6] proposed a novel Bayesian-based model to investi­

gate the collision risk between ships and offshore wind farms. Moreover,
Ship collision is the most frequently occurring accident and much there are also many studies analyzing the probability or severity of ship
attention has been attracted to prevent such accident worldwide. Zhang collisions, such as reducing the probability of ship collisions by adjusting
et al. [1] proposed a quantitative methodology for maritime risk the passage schedule in the Istanbul Strait [7], estimating multi-ship
assessment to identify the key factors contributing to collisions in the collision probability in Victoria Harbor [8], evaluating the probability
Tianjin port. Cheng et al. [2] presented a fuzzy-logic-based model for of an icebreaker–ship collision in an ice channel from the Arctic area [9],
evaluating ship collision risks in constrained crossing scenarios. Yu et al. predicting the severity of ship collision accidents based on accident data
[3] considered spatial-temporal factors to assess collision risk for pre­ [10,11].
venting accidents in the coastal waters of North China, Korean, and As a special type of collision accident, the collisions between mer­
Japan. Mazurek et al. [4] proposed a novel method for identifying chant ships and fishing vessels are pronounced in China because of the
collision-prone locations for oil tankers in the Gulf of Finland. Cai et al. fishery economy. Based on the data from the National Bureau of Sta­
[5] assessed the collision risk of ferries to enhance navigation safety tistics of China and the Annual Chinese Fishery Economic Statistical
based on Automatic Identification System (AIS) data in the Yangtze Bulletin in Fig. 1, it can be seen that the gross output value of the fishery

* Corresponding author.
E-mail address: [email protected] (B. Wu).

https://doi.org/10.1016/j.ress.2024.110201
Received 16 August 2023; Received in revised form 11 April 2024; Accepted 13 May 2024
Available online 14 May 2024
0951-8320/© 2024 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
H. Wang et al. Reliability Engineering and System Safety 249 (2024) 110201

economy continuously increased from 296.22 billion in 2013 to 472.58 Table 1

billion USD in 2022. Owing to the large number of fishing vessels Some typical studies on the identification of influencing factors.
navigating close to the coast, collision accidents between merchant ships Study Accident type Techniques applied Research topic
and fishing vessels have occurred frequently, causing the sinking of
[27] Collision HFACS and fault tree Analyze risk factors for
fishing vessels or loss of life and property [12,13]. Take the accident accidents the collision during
statistics of the Liaoning Maritime Safety Administration as an example, icebreaker assistance.
the most serious consequences of the accident were the collisions be­ [28] Collision and Geographic information Propose a causal network
tween merchant ships and fishing vessels from 2017 to 2021. grounding system, HFACS and to assess the risk of
accidents Bayesian network maritime accidents.
Different from the collisions between merchant ships, the collisions [29] Maritime HFACS, fault tree analysis, Analyze the human
between merchant ships and fishing vessels have some distinguishing accidents fuzzy theory, artificial factors of maritime
characteristics. First, merchant ships and fishing vessels are overseen by neural network accidents.
different authorities. Specifically, merchant ships are overseen by the [30] Marine HFACS-PV Identify human and
accidents organizational factors of
Ministry of Transport while fishing vessels are overseen by the Ministry
marine accidents.
of Agriculture and Rural Affairs in China. There is a difference in the [31] Collision HFACS, decision-making Derive relationships
laws, regulations, and technical standards of ship management, which accidents trial and evaluation between human factors
may make the merchant and fishing vessels different in their ship laboratory, and fuzzy for collision accidents.
behavior. For instance, although both merchant and fishing vessels cognitive map

should obey the Convention on the International Regulations for Pre­

venting Collisions at Sea, 1972 (COLREGs), the crews on the fishing obtained from sensitivity analysis [24,25] and the development paths to
vessels seem to have a larger probability of not obeying the COLREGs causing accidents can be derived through the strength of influence
owing to a lack of basic safety training [14]; among the 185 collision analysis [26].
accidents between merchant ships and fishing vessels, fishing vessels are Therefore, the motivation of this paper is to develop a hybrid model
involved in illegal fishing in 68 cases. Second, the equipment on the combining HFACS with BN to further analyze the HOFs of collision ac­
fishing vessels and merchant ships is also quite different. Merchant ships cidents between merchant ships and fishing vessels, and to identify the
use Very High Frequency (VHF) to communicate with each other while key influencing factors and the development paths for the prevention of
fishing vessels use shortwave radio in China. collision accidents. The remainder of this paper is organized as follows.
Although the collisions between merchant ships and fishing vessels Section 2 proposes the methodology for the human and organizational
are special, there are few studies managed to develop a quantitative factor analysis of collision accidents. Section 3 quantifies the HFACS-BN
model to identify key influencing factors within the human and orga­ model and validates the proposed method. Section 4 discusses limita­
nization factors (HOFs), which are the prominent causation factors for tions of the proposed model. Conclusions are drawn in Section 5.
collision accidents between merchant ships and fishing vessels. For
maritime accidents, Human factors analysis and Classification System 2. Methodology for human and organizational factors analysis
(HFACS) is widely used because it provides a four-layer structure for
categorizing HOFs in accident scenarios [15], and some typical studies 2.1. Establish the human and organizational factors analysis framework
on HOFs are summarized in Table 1. As the HFACS method cannot
quantify the HOFs, Bayesian Network (BN), which can quantify the The HFACS method, which provides a hierarchical framework for
relationship between HOFs from a probability perspective [16-21], is analyzing the human and organizational factors, is introduced to iden­
introduced in the HOFs analysis of collision accidents between merchant tify the influencing factors for collisions between merchant ships and
ships and fishing vessels in this paper owing to the following reasons. fishing vessels. Moreover, the Bayesian network, which can graphically
First, the graphical structure of BN can represent the causal relationship and quantitatively describe the relationships of the influencing factors,
between the influencing factors [22]. Second, it can describe the process is integrated to derive the key HOFs and the development paths.
of accident development [11,23]. Third, key influencing factors can be Therefore, the proposed HOFs analysis framework for collision accidents

Fig. 1. The gross output value of the fishery economy.

2
H. Wang et al. Reliability Engineering and System Safety 249 (2024) 110201

between merchant ships and fishing vessels is established in three steps, 2.2. Human and organizational factors identification using modified
which is shown in Fig. 2. HFACS

Step 1. Establish the HFACS framework to identify the influencing 2.2.1. Modifying the human factors analysis and classification system
factors. The modified HFACS framework is developed for qualitative framework
human and organizational factor analysis of collision accidents be­ Based on the Swiss Cheese Model [32], the HFACS method was
tween merchant ships and fishing vessels. The influencing factors are developed by Shappell and Wiegmann [33] to investigate the direct and
identified at each level of HFACS based on statistical data from 443 indirect human causes of aviation accidents. The HFACS is widely used
collision accidents which are primarily caused by HOFs, previous for identifying HOFs in maritime accidents [30,31] owing to the three
studies, and expert judgments. distinguishing features. First, it provides a rational and comprehensive
Step 2. Develop the HFACS-BN model for quantitative analysis. The hierarchical classification system to identify the HOFs in maritime ac­
graphical structure of the Bayesian network model is formed by the cidents [34]. Second, it is possible to discover the underlying causal
transformation of the modified HFACS. The marginal probabilities factors of accidents [28]. Third, the model is flexible to be modified
and conditional probability tables (CPTs) are derived by parameter according to different accident types [15].
learning with Expectation Maximization (EM) algorithm. The original HFACS is divided into four levels, which are organiza­
Step 3. Model validation and results analysis. Two axioms are tional influences, unsafe supervision, preconditions for unsafe acts, and
introduced to validate the proposed HFACS-BN model. Sensitivity unsafe acts [35,36]. In this paper, the original HFACS framework is
analysis is introduced to identify the key HOFs for target nodes, and modified to extract accident causal factors of collisions between mer­
the strength of influence is calculated to obtain the development chant ships and fishing vessels, and HOFs are classified into five cate­
paths. gories including external factors, organizational influences, unsafe
supervision, preconditions for unsafe acts, and unsafe acts [37]. More­
over, the influencing factors are also slightly different from the original
HFACS owing to the distinguishing characteristics of this type of

Fig. 2. The research framework of human and organizational factors analysis.

3
H. Wang et al. Reliability Engineering and System Safety 249 (2024) 110201

collision. The five levels are defined as follows. (1) External factors factor is considered because the safety management department of the
(Level 1). This level does not directly contribute to collision accidents fishing vessels is different from merchant ships, which may also facili­
but influences organizational factors. It includes insufficient regulation tate collision accidents between merchant ships and fishing vessels [38].
and enforcement, imperfect management mechanisms, and inadequate (3) Unsafe supervision (Level 3). This level includes inadequate super­
information organization. Inadequate regulations and poor enforcement vision, inappropriate operations, and failure to correct problems. As
may cause inadequate safety management on merchant ships and fishing supervisory violations and inadequate supervision are difficult to
vessels, or non-compliance with rules and regulations, which increases distinguish directly from accident investigation reports, supervisory
the hazards of collision accidents. Imperfect management mechanisms violations are integrated into inadequate supervision in this proposed
are related to the maritime safety administration and fisheries man­ model. (4) Preconditions for unsafe acts (Level 4). This level includes
agement departments. Information technology is introduced to improve coordination factors, personnel factors, and environmental factors. Co­
maritime safety, while inadequate information technology may influ­ ordination factors differ from the original HFACS owing to the coordi­
ence the information exchange between merchant ships and fishing nation of merchant ships and fishing vessels is difficult because fishing
vessels, which is a predominant factor for collision accidents between vessels intend to catch fish while merchant ships intend to sail to the
merchant ships and fishing vessels in China because the merchant ships destination. Moreover, the maneuverability of the merchant ships and
use the VHF while the fishing vessels use shortwave radio. (2) Organi­ fishing vessels are also different, which may cause the situation
zational influences (Level 2). This level includes resource management, awareness of these two types of ships to be quite different. (5) Unsafe
safety management, and organizational climate and process. Safety acts (Level 5). This level includes judgment errors, skill-based errors,
management is considered at this level because it can prevent collision and violations [39]. The overview of the modified HFACS framework for
accidents, which is a new factor compared with the original HFACS. This analyzing HOFs of collision accidents between merchant ships and

Fig. 3. The modified HFACS framework. Adapted from [40].

4
H. Wang et al. Reliability Engineering and System Safety 249 (2024) 110201

fishing vessels is shown in Fig. 3. from phrases like “Listening to music while on duty”, “fatigue driving”
Moreover, the reason of utilizing HFACS to identify HOFs of colli­ or “engaging in off-task activities”, indicating that crews are subject to
sions between merchant ships and fishing vessels is summarized as fol­ physical or mental limitations while working. The typical keywords and
lows. First, HFACS, as a widely used HOFs analysis framework, has been sentences for identifying the influencing factors are given in Table A1 of
applied to various types of maritime accidents [27-31]. It has been Appendix A.
modified to investigate collisions between merchant ships and fishing Third, to identify the special influencing factors for distinguishing
vessels in this paper. collisions between merchant ships and fishing vessels from other types
Second, when identifying traditional and special factors, both mer­ of ship collisions, the influencing factors are identified in two steps.
chant ships and fishing vessels can utilize the HFACS framework to First, the influencing factor is determined for both the merchant ship
analyze HOFs of ship collisions, including five categories, which are and the fishing vessel. Take the factor “Unsafe speeds” as an example, it
external factors, organizational influences, unsafe supervision, pre­ is defined as an influencing factor if the fishing vessel or the merchant
conditions for unsafe acts, and unsafe acts. Fig. 4 presents the identifi­ ship violated the unsafe speed. Second, the special influencing factors
cation of HOFs for collisions involving merchant ships and fishing categorized as merchant and fishing vessel collisions are identified
vessels in the HFACS framework. through the factors that are primarily violated by fishing vessels and the
characteristics of merchant and fishing vessel collisions.
2.2.2. Identifying the human and organizational factors Finally, the influencing factors are classified into sub-factors of the
To further identify the HOFs of collision accidents, the modified HFACS framework by five experts, and the information of the experts in
HFACS method, together with the previous studies and expert judg­ detail is shown in Table 2. The hierarchical structure of the influencing
ments, are integrated to obtain the influencing factors. The process of factors is developed based on the HFACS framework considering the
identifying HOFs is shown in Fig. 5. previous studies [31,52] and expert judgments.
First, to identify the HOFs, accident investigation reports of 201 The identified influencing factors and explanations are given in the
collisions between merchant ships and fishing vessels, and 262 collisions tables. The frequency of these factors is listed in the fifth column, and the
between merchant ships, are collected from the Maritime Safety method used to identify them is indicated in the last column, with “E”
Administration of the People’s Republic of China [41] and its subordi­ representing expert judgments and “A” representing the accident data­
nate maritime bureaus [42-51]. After eliminating 20 reports lacking key base. In Bayesian network modeling, abbreviations including “S” are
information on HOFs, 185 accident investigation reports on collisions used to describe the special influencing factors, as indicated by the
between merchant ships and fishing vessels and 258 accident investi­ symbols in the third column of the table. 11 special influencing factors
gation reports on collisions between merchant ships during the period are identified by this method, including insufficient regulation and
from 2013 to 2023, have been selected. enforcement (E1-S), imperfect management mechanisms (E3-S), inade­
Second, the influencing factors of HOFs are identified based on the quate manning (OR1-S), lack of effective administration of fishing ves­
accident investigation reports. The identification of influencing factors sels (OO1-S), communication barriers (OO3-S), insufficient auditing of
in each accident investigation report is carried out by keyword mining or fishing vessel certificates (UI1-S), fishing methods (UF2-S), unaware of
sentence mining. Two types of influencing factors are identified. The regulations and legislation (UP3-S), insufficient communication (PC1-
first type is identified through direct keyword mining or sentence min­ S), illegal fishing (AV1-S), and crossing the shipping lane (AV3-S).
ing, such as “Improper lookout”, which could be derived from phrases
like “negligence in lookout” or “absence of a formal lookout”. The sec­ (1) Level 1 (External factors). This level includes three influencing
ond type is identified through indirect keyword mining or sentence factors, which are insufficient regulation and enforcement,
mining, such as “Physical or mental limitations”, which could be derived inadequate information organization, and imperfect

Fig. 4. Identify the HOFs for two types of collisions using the HFACS framework.

5
H. Wang et al. Reliability Engineering and System Safety 249 (2024) 110201

Fig. 5. The process of identifying human and organizational factors.

management mechanisms. Specifically, the influencing factors of

Table 2
collision accidents in Level 1 are shown in Table 3.
The information of the five experts.
(2) Level 2 (Organizational influences). This level includes three
Expert Institution Age Education Experience types of influencing factors, which are resource management,
No.
safety management, and organizational climate and process.
E1 University 43 Doctor He focuses on the study of Moreover, resource management includes inadequate manning,
degree maritime accident
emergency management, and insufficient facilities and equip­
analysis and prevention.
E2 University 50 Doctor He focuses on the study of
ment; organizational climate and process include lack of effective
degree ship navigation administration of fishing vessels, failure to implement regulations
technology. or rules, and communication barriers; safety management in­
E3 Maritime Safety 42 Master He was involved in more cludes insufficient navigation safety education and training, poor
Administration degree than 100 maritime
safety culture, and insufficient safety assessment technology.
accidents analysis,
including at least 20 Specifically, the influencing factors of collision accidents in Level
collision accidents 2 are shown in Table 4.
between merchant ships (3) Level 3 (Unsafe supervision). This level includes three types of
and fishing vessels.
influencing factors, which are inadequate supervision, failure to
E4 Shipping company 50 Bachelor He is a captain on the
degree ocean-going ship and has
correct known problems, and inappropriate operations. Further­
25 years of experience in more, inadequate supervision includes insufficient auditing of
ship navigation. fishing vessel certificates, inadequate assessment of education
E5 Ministry of 53 Master He focuses on the and training, and personal factors of the supervisor; failure to
agriculture and degree supervision of fishing
correct known problems includes insufficient risk assessment,
rural affairs vessels and has conducted
several projects on the fishing methods, and insufficient safety reporting; inappropriate
safety of fishing vessels in operations include unreasonable navigation plan, insufficient
Rizhao Waterways. supervision plan, and unaware of regulations and legislation. The
influencing factors of collision accidents in Level 3 are shown in
Table 5.

Table 3
Influencing factors of collision accidents (Level 1).
Level 1 Influencing factors Symbol Explanations Frequency E=Expert judgments,
A=Accident database

External Insufficient regulation E1-S Owing to weak administrative law enforcement, issues such as the unseaworthiness 172 A& [53]
factors and enforcement of ships, unsuitable crews, inadequate shipping company safety management, and
poor organizational management persist, potentially leading to unsafe ship
navigation.
Inadequate information E2 This factor will influence the information exchange both between ships and 73 E& A
organization between ship and shore.
Imperfect management E3-S As the safety management of merchant ships is overseen by the maritime safety 89 E& A
mechanisms administration, while the safety management of fishing vessels is overseen by the
fishing management departments, it is very hard to have a coordinated policy to
enhance the safety of merchant ships and fishing vessels.

6
H. Wang et al. Reliability Engineering and System Safety 249 (2024) 110201

Table 4
Influencing factors of collision accidents (Level 2).
Level 2 Influencing factors Symbol Explanations Frequency E=Expert judgments,
A=Accident database

Resource Inadequate manning OR1-S Inadequate crew numbers, expired seafarer certification, unfit 244 A& [54]
management certificate matches, or a lack of certificates that meet specific
navigational area requirements.
Emergency management OR2 Lack of emergency training, resources, or preparedness for merchant 160 E& A & [11]
ships and fishing vessels.
Insufficient facilities and OR3 Insufficient communication equipment, navigation equipment, and 76 A& [30]
equipment onboard safety equipment.
Organizational Lack of effective OO1-S The recruitment, education, and assessment of crews for fishing 91 E& A& [55]
climate and process administration of fishing vessels are not rigid, and risk management for fishing vessels is
vessels inefficient.
Failure to implement OO2 Failure to implement regulations or rules related to safe navigation, 175 A
regulations or rules such as safety management system, STCW convention.
Communication barriers OO3-S Failure to understand communications terminology, the different 121 A& [56]
communication equipment, and adverse working conditions.
Safety management Insufficient navigation safety OS1 Insufficient training in sailing skills, navigation regulations, the 313 A& [57]& [58]
education and training understanding and implementation of collision avoidance rules, and
the use of navigation equipment.
Poor safety culture OS2 The companies don’t develop a “safety first” culture. 246 E& A& [59]
Insufficient safety OS3 Inadequate collision alert, monitoring, and navigation systems. 232 A& [38]
assessment technology

Table 5
Influencing factors of collision accidents (Level 3).
Level 3 Influencing factors Symbol Explanations Frequency E=Expert judgments,
A=Accident database

Inadequate Insufficient auditing of UI1-S Ineffective supervision of fishing vessel certificates by fisheries 48 A& [56]
supervision fishing vessel certificates authorities or fishing vessel operators.
Inadequate assessment of UI2 Inadequate examination of knowledge and practical operation of 238 A& [55,60]
education and training professional skills for crew members.
Personal factors of the UI3 Inadequate risk assessment capacity or duty fulfillment. 122 E& A
supervisor
Failure to correct Insufficient risk UF1 Failure to identify hazards due to lack of knowledge, experience, skills, 239 E& A
known problems assessment equipment, etc.
Fishing methods UF2-S With a range of methods utilized by fishing vessels, including bait fishing, 58 A& [61,62]
fly-fishing, bait casting, spinning, and trolling, the judgment capacities of
merchant vessels could potentially be impacted.
Insufficient safety UF3 Failure to report unsafe issues or vessel port entries and exits. 122 A
reporting
Inappropriate Unreasonable navigation UP1 Improper route planning, non-compliance with regulations, misjudgment 165 E& A
operations plan on ship’s capability.
Insufficient supervision UP2 The lack of a well-structured supervision plan. 212 E& A
plan
Unaware of regulations UP3-S The comprehension of regulations and legislation may not be complete 94 E& A& [54]
and legislation among captains and crews of fishing vessels.

Table 6
Influencing factors of collision accidents (Level 4).
Level 4 Influencing factors Symbol Explanations Frequency E=Expert judgments,
A=Accident database

Personnel factors Insufficient experience and PP1 Lack of awareness, experience, and skills of crew members for 343 A
skills merchant ships and fishing vessels.
Unfamiliarity with collision PP2 Failure to understand COLREGs and the inland waterway collision 226 E& A
avoidance rules regulations of the People’s Republic of China.
Physical or mental limitations PP3 The fatigue, alcohol abuse, disease of the crews, and insufficient 73 A& [12]
emergency response capacity.
Coordination Insufficient communication PC1-S A lack of, or ineffective, external, and internal communication. 272 A
factors Inadequate use of sound or PC2 Sound and light signals are not used correctly or effectively as 132 A
light signals required by rules.
Improper duty PC3 The staff on duty is not organized properly or fails to fulfill their 178 A
safety responsibilities.
Environmental Restricted visibility PE1 This may influence the perception of the surrounding 73 A& [63-65]
factors environment and the dynamics of other vessels.
Complex navigational PE2 Congested waterways, narrow navigation channels, fishing areas, 249 A& [66]
environment and other complex conditions.
Meteorological and PE3 Strong winds, big waves, fast water flow, and other bad hydro- 143 A& [67,68]
hydrological conditions meteorological factors.

7
H. Wang et al. Reliability Engineering and System Safety 249 (2024) 110201

(4) Level 4 (Preconditions for unsafe acts). This level includes three process. (1) HOFs at higher levels in HFACS are influenced by lower
types of influencing factors, which are personnel factors, coor­ levels [72]. (2) Five levels excluding external factors have a direct in­
dination factors, and environmental factors. Moreover, personnel fluence on the occurrence of the collision [73,74]. (3) HOFs at the same
factors include lack of experience and skills, unfamiliarity with level in HFACS are independent of each other, and the parent nodes only
collision avoidance rules, and physical or mental limitations; directly influence the associated child nodes [75]. Based on these as­
coordination factors include insufficient communication, incor­ sumptions, the graphical structure of the BN model is determined, and
rect use of sound or light signals, and improper duty; environ­ the principle of mapping HFACS into BN is given in Fig. 6.
mental factors include restricted visibility, complex navigational The developed graphical structure of BN, which is shown in Fig. 7,
environment, and meteorological and hydrological conditions. involves 56 factors related to collision accidents. In this figure, the
The influencing factors of collision accidents in Level 4 are shown different influencing factors at each level in HFACS are represented in
in Table 6. different colors, and the relationships between each level are defined
(5) Level 5 (Unsafe acts). This level includes three types of influ­ according to the relationships in HFACS.
encing factors, which are judgment errors, skill-based errors, and
violations. Moreover, judgment errors include misjudgment on 2.4. Determining the quantitative relationships
collision risk, incorrect use of navigational aids, and insufficient
lookout; skill-based errors include improper actions for collision After determining the dependencies between the nodes to develop
avoidance, failure to detect other vessels, and unsafe speeds; vi­ the BN structure, the marginal probabilities for each node and condi­
olations include illegal fishing, illegal navigation of merchant tional probability tables for the intermediate and leaf nodes should be
ships, and crossing the shipping lane. The influencing factors of calculated to quantify the influence of the nodes [76]. To obtain BN
collision accidents in Level 5 are shown in Table 7. parameters based on statistical data, EM algorithm, Maximum Likeli­
hood Estimation, and Bayesian estimation are widely used [77]. From
the accident investigation reports, missing values are detected in some
2.3. Developing the graphical structure by mapping HFACS into BN
cases, which means incomplete data exists from the accident database.
EM algorithm, which can estimate parameters from incomplete data
Bayesian network is widely used for risk assessment of maritime
[78], is introduced for parameter learning of BN. EM algorithm,
transportation owing to the qualitative and quantitative representation
including the expectation step (E-step) and the maximization step
of the relationships among a variety of influencing factors. To quantify
(M-step), is described as follows [79].
the HOFs of collision accidents between merchant ships and fishing ( ) (
vessels, the HFACS is mapped into the BN, which is carried out in two The incomplete dataset is defined as D = l1 ,l2 ⋯lM , while Y = c1 ,
)
steps. c2 ⋯cM is the complete dataset, D ∈ Y. Let ς = (V, E, P) be a BN with
The first step is to determine the nodes of the BN and their states. The { } { } {
parameters Θ = Θq , where Θq = Θqs and Θqs = Θqst such that
}
root and intermediate nodes are transformed from the identified HOFs in ( ⃒ ( ) )
Θqst = P Xq = t ⃒pa Xq = s for each q, s, t.
each layer of HFACS. The leaf node is defined as collision, which is the
result of a mixture of factors. The state of each node is determined based K
∑
on the statistical data in the accident investigation report. Taking k(Θ) = logP(Xk |Θ) (1)
k=1
“Communication barriers” as an example, the node state is represented
by “Existence/Nonexistence” to indicate whether there is a communi­ M
( ) ∑ ( ⃒ )
cation barrier in the collision between the merchant ship and the fishing EΘ Mqst = P Xq = t, pa(Xi ) = s⃒lq , Θq , D (2)
vessel. Specifically, the state of “Collision” is denoted by “Collision1” or q=1

“Collision2”, where “Collision1” indicates collisions between merchant

First, the E-step is to establish a complete data set Y using Eqs. (1)-
ships and fishing vessels, and “Collision2” indicates collisions between ( ( ))
(2). Where Mqst represents the number of Xq , pa Xq = (t, s) and lq
merchant ships.
represents the qth case of D.
The second step is to determine the relationship between each node.
The relationships between HOFs are based on the relationships of each
level in the HFACS framework. Three assumptions are defined in this

Table 7
Influencing factors of collision accidents (Level 5).
Level 5 Influencing factors Symbol Explanations Frequency E=Expert
judgments,
A=Accident
database

Judgment Misjudgment on collision AJ1 Failure to accurately assess and estimate the risk of collision with other vessels. 315 A
errors risk
Incorrect use of AJ2 Improper use or non-use of navigational equipment such as bridge radar, AIS, VHF. 196 A
navigational aids
Insufficient lookout AJ3 Failing to use all effective means to ensure regular watching. 385 A& [69]
Skill-based Improper actions for AS1 Failed to take timely, effective collision avoidance measures based on collision 393 A
errors collision avoidance avoidance rules or to use good seamanship for careful navigation.
Failure to detect other AS2 Failure to detect or delay in detecting other vessels threatening to collide. 234 A& [55] & [70]
vessels
Unsafe speeds AS3 Failure to adjust navigation speed for safety in varying maritime conditions or 183 A& [10,71]
hazards.
Violations Illegal fishing AV1-S Engaging in fishing activities in port channels, anchorage areas, during closed 68 A& [53]
fishing seasons, as well as unlicensed fishing.
Illegal navigation of AV2 Overloading of merchant ships, traversing beyond designated shipping routes, or 74 A
merchant ships navigating within restricted areas.
Crossing the shipping lane AV3-S Improper navigation of fishing vessels through and across shipping lanes or deep- 21 A
water channels.

8
H. Wang et al. Reliability Engineering and System Safety 249 (2024) 110201

Fig. 6. The principle of mapping HFACS into Bayesian networks.

Fig. 7. The graphical structure of the Bayesian network model.

( )
EΘ Mqst two successive iterations does not exceed the log-likelihood of threshold
θ∗qst = ∑‖Xq ‖ ( ) (3) δ in Eq. (4). where kq (Θ) is the log-likelihood of Θ after the qth iteration
s=1 EΘ Mqst
and kq+1 (Θ) is the log-likelihood of Θ after the (q + 1)th iteration.
Second, the M-step takes the new estimates θ∗qst of θqst in the sufficient
data from the complete data of cases using Eq. (3).
2.5. Sensitivity analysis and the strength of influence analysis
⃒ ⃒
kq (Θ) − kq+1 (Θ)⩽δ⃒kq+1 (Θ)⃒ (4)
2.5.1. Sensitivity analysis for identifying the key human and organizational
The iteration of these two steps continues until convergence of k(Θ) factors
is achieved. This occurs when the variation in log-likelihoods between Sensitivity analysis can be introduced to identify the key contrib

9
H. Wang et al. Reliability Engineering and System Safety 249 (2024) 110201

uting factors [80]. In this paper, the sensitivity analysis is carried out by 2.5.2. The strength of influence analysis for deriving the development paths
calculating the derivatives of the posterior probabilities of the target The strength of influence which represents the impact on the child
variables of the BN for different nodes [81]. Define x as the probability node by the parent node is calculated based on the CPTs of the child
that a node of the variable is in a specific state. Based on different pieces node and expresses the distance between various conditional probability
of evidence, the posterior probability of the target node can be expressed distributions over the child node, given the states of the parent node
as a function of x, as shown in [82]. The Hellinger distance is one of the methods used to calculate the
distance between distributions. It has better sensitivity and uses a value
αx + β
f(x) = (5) higher than 0 and lower than 1 to represent the relative differences [83].
γx + δ
{ ⃒ }
⃒ ∑n
⃒
where α, β, γ, δ are constants to x and can be obtained by the method in P, Q ∈ (p1 , p2 , ⋯, pn )⃒pi > 0, pi = 1 , n > 1 (7)
⃒
[81]. i=1

The derivative of the posterior probability of the target node in a √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅

∑n ( √̅̅̅̅ √̅̅̅̅)2
specific state can be obtained by taking the derivative of f(x) to x, as H(P, Q) = pi − qi (8)
i=1
shown in
H(P, Q)
ʹ
f(x) =
αδ − βγ
(6) Hnorm (P, Q) = √̅̅̅ (9)
2 2
(γx + δ)
Sensitivity analysis is used to identify the key HOFs on the risk of P and Q are two discrete probability distributions and Eq. (7) represents
collision accidents. A higher value of the parameter derivative indicates the definition of P and Q.
that the parameter has a greater impact on the target node. Note that the The Hellinger distance is calculated in Eq. (8). The result is
derivative values calculated using the sensitivity analysis formula normalized using Eq. (9). In this way, the strength of influence between
depend on the collected dataset. If the dataset of collision accidents all child nodes and parent nodes in BN can be derived. Sequentially
increases, the derivative value will need to be updated. The results can connecting each child node to the parent node with the highest influ­
be a reference for the maritime safety authorities to enhance their safety ence, the development path of the target node can be obtained [26].
management.

Fig. 8. The parameter learning results of the HFACS-BN model.

10
H. Wang et al. Reliability Engineering and System Safety 249 (2024) 110201

3. Application of the HFACS-BN model for human and Table 8

organizational factors analysis The conditional probabilities of “coordination factors”.
Insufficient Inadequate use of Improper Coordination
3.1. Quantification of the proposed HFACS-BN model communication sound or light signals duty factors

Bad Good
After introducing EM algorithm for parameters learning, the HFACS-
Yes Yes Yes 0.9149 0.0851
BN model for the collision accidents can be developed and the results are Yes Yes No 0.7778 0.2222
shown in Fig. 8. It can be seen that the state one of “Collision” with a Yes No Yes 0.7931 0.2069
probability of 41.87%, while the probability for the state two is 58.13%. Yes No No 0.6989 0.3011
When setting the collision state one as 100%, it means collisions be­ No Yes Yes 0.6923 0.3077
No Yes No 0.3333 0.6667
tween merchant ships and fishing vessels, while setting state two as
No No Yes 0.5806 0.4194
100%, it denotes collisions between merchant ships. Fig. 9 illustrates the No No No 0.0300 0.9700
comparison of the influence of the four HOFs on these two types of
collisions. The comparison indicates that collisions between merchant
ships and fishing vessels are more influenced by organizational in­ probability of a child node. This axiom is verified by changing the
fluences, unsafe supervision, and preconditions for unsafe acts than probability between the child node “Organizational influences” and its
collisions between merchant ships, while collisions between merchant parent nodes. As can be seen from Table 9, taking “Organizational in­
ships and fishing vessels are less influenced by unsafe acts than collisions fluences” as an example, if the probability for the “Inadequate” state of
between merchant ships. Among the four HOFs of collisions, unsafe acts the parent node “Safety management” increases from 71% to 100%,
are the most contributing factors for two types of collisions. Moreover, then the probability for the “Bad” state of the child node “Organizational
improper actions for collision avoidance (AS1) account for 88.71%, influences” increases from 59% to 65%; if the probability for the
which indicates that this factor is the most frequently occurring factor “Inadequate” state of the parent node “Safety management” decreases
among all HOFs for collision accidents. from 71% to 0%, the probability for the “Bad” state of the child node
Moreover, the CPTs of all intermediate nodes and the leaf node are “Organizational influences” reduces from 59% to 45%. The effect of the
derived through EM parameter learning. The conditional probabilities
for the coordination factors as an example are shown in Table 8.
Table 9
Validation of Axiom 1 for the “Organizational influences” node.
3.2. Validation of the proposed HFACS-BN model Parent nodes and the states Probabilities for Probabilities of
parent nodes “Organizational influences
The accuracy of the BN model is verified by satisfying two axioms (Bad)”

[84]. External factors (Bad) 0.46 0.59

1 0.72
Axiom 1. The posterior probability of the child node should be rela­ 0 0.48
tively increased/ decreased when there is a slight increase/decrease in Resource management 0.55 0.59
the prior probability of each parent node. (Insufficient) 1 0.66
0 0.51
Organizational climate and 0.39 0.59
Axiom 2. Although the variation of the probability distributions of process (Imperfect) 1 0.66
each parent node, the magnitude of its influence on the values of the 0 0.55
child node should remain consistent. Safety management 0.71 0.59
(Inadequate) 1 0.65
According to Axiom 1, an increase or decrease in the probability of a 0 0.45
parent node will cause a corresponding increase or decrease in the

Fig. 9. The influence of the four HOFs on two types of collisions.

11
H. Wang et al. Reliability Engineering and System Safety 249 (2024) 110201

other parent nodes on the child node can also be verified using the same factors increase the collision risk are due to third perspectives. First,
method [59]. The results of the verification indicate that the proposed crews lacking experience or training are unable to accurately evaluate
HFACS-BN model satisfies Axiom 1. the situation of the vessel, which will cause misjudgments on the safety
Axiom 2 is proved by verifying the probability variation between situation. Second, crews unfamiliar with collision avoidance rules will
“Judgment errors” and its parent nodes. Eleven pieces of evidence are make the other ships not understand the intention, which will increase
assigned to the parent nodes “Misjudgment on collision risk (AJ1)”, the collision risk. Third, physical or mental limitations have a significant
“Incorrect use of navigational aids (AJ2)” and “Insufficient lookout impact on operations and response times.
(AJ3)”, respectively. The posterior probabilities of the “Judgment er­ Unsafe supervision is the third sensitive level for collision accidents,
rors” node are derived. The probability distribution for the “Yes” state of among which failure to correct known problems is the most sensitive
the “Judgment errors” node based on its three parent nodes is shown in influencing factor for collision accidents at this level, followed by
Fig. 10, which shows that the change in the probability of “Judgment inappropriate operations, and inadequate supervision. The main reason
errors” is consistent with the change in the probability of “AJ1”, “AJ2” is that many pre-existing problems such as outdated or inadequate
and “AJ3”. This means that the probability of the child node increases as fishing equipment or methods, insufficient risk assessment and safety
the probability of the parent node increases, with a consistent trend and reporting are not addressed leading to the risk of collision accidents.
no unusual fluctuations [28,59]. Other relationships between parent Moreover, there are many self-operated fishing vessels involved in the
nodes and the child node can also be verified in the same method. The accident, which do not have their supervisory organizations.
results of the verification indicate that the proposed HFACS-BN model Organizational influences are the least sensitive level for collision
satisfies Axiom 2. accidents but still pose a risk to the collision. Safety management is
sensitive to the organizational influences on collision accidents at this
level, followed by resource management, and organizational climate
3.3. Identifying the key human and organizational factors and processes. Safety management is important owing to several as­
pects. First, both crews on merchant and fishing vessels require training
Define “Collision” as the target node, the derivatives of its posterior and expertise to effectively handle the challenges during navigation.
probability to other nodes are calculated using Eqs. (5)-(6), and the Second, the safety culture of both merchant and fishing vessels can
obtained results are shown in Fig. 11. The results show that the most significantly impact the behavior of crew members. Finally, risk
sensitive level for collision accidents is unsafe acts, followed by pre­ assessment plays a crucial role in identifying potential safety hazards
conditions for unsafe acts, unsafe supervision, organizational influences, and adopting effective safety measures, which is significant in miti­
and external factors. Unsafe acts are the most sensitive influencing gating the likelihood of collision accidents between merchant ships and
factors in collision accidents owing to two perspectives. Crews on fishing fishing vessels.
vessels with limited navigational experience and inadequate collision Moreover, sensitivity analysis is introduced to identify key influ­
avoidance awareness may take unsafe acts such as negligent lookouts for encing factors to collisions. Table 10 gives detailed information on the
fishing operations, and non-compliance with collision avoidance rules, top 15 influencing factors, with the bolded parts highlighting the special
making it difficult to prevent collision accidents. Some merchant ships influencing factors of collisions between merchant ships and fishing
take unsafe acts, such as neglecting to properly monitor their sur­ vessels. This finding is useful because special attention should be paid to
roundings, unsafe speeds, and improper collision avoidance actions. these factors by the Maritime Administration, fisheries supervisory
These actions increase the risk of collision accidents. agencies, shipping companies, and fishermen to prevent collision
Preconditions for unsafe acts are the second most sensitive level for accidents.
collision accidents. Specifically, personal factors are the most sensitive The modified HFACS includes five levels, where influencing factors
influencing factors to collision accidents at this level, followed by co­ at the lower levels directly influence the occurrence of influencing fac­
ordination factors, and environmental factors. Personal factors include tors at the upper levels and finally influence the occurrence of collision
lack of experience and skills, unfamiliarity with collision avoidance accidents. It is essential to identify the key influencing factors at each
rules, and physical or mental limitations. The reasons why personal

Fig. 10. The probability distribution for the “Judgment errors” node.

12
H. Wang et al. Reliability Engineering and System Safety 249 (2024) 110201

Fig. 11. Sensitivity analysis results of “Collision”.

Table 10 Table 11
Top 15 influencing factors to “collision”. The key influencing factors in each level based on sensitivity analysis.
The influencing factors (The root Symbol Average of sensitivity Rank The target nodes The key influencing factors in each level (The root nodes)
nodes) analysis
External factors Insufficient regulation and enforcement (E1-S),
Insufficient regulation and E1-S 8.93E-3 1 Inadequate information organization (E2), Imperfect
enforcement management mechanisms (E3-S)
Improper actions for collision AS1 8.50E-3 2 Organizational Insufficient navigation safety education and training
avoidance influences (OS1), Inadequate manning (OR1-S), Insufficient safety
Insufficient navigation safety OS1 7.33E-3 3 assessment technology (OS3), Failure to implement
education and training regulations or rules (OO2), Insufficient facilities and
Inadequate information organization E2 6.51E-3 4 equipment (OR3)
Inadequate assessment of education UI2 6.14E-3 5 Unsafe supervision Insufficient supervision plan (UP2), Inadequate
and training assessment of education and training (UI2), Insufficient
Insufficient supervision plan UP2 5.98E-3 6 risk assessment (UF1), Fishing methods (UF2-S),
Insufficient risk assessment UF1 5.87E-3 7 Insufficient auditing of fishing vessel certificates (UI1-S)
Inadequate manning OR1-S 5.85E-3 8 Preconditions for Insufficient communication (PC1-S), Unfamiliarity with
Imperfect management mechanisms E3-S 5.79E-3 9 unsafe acts collision avoidance rules (PP2), Insufficient experience
Insufficient communication PC1-S 5.58E-3 10 and skills (PP1), Complex navigational environment
Fishing methods UF2-S 5.12E-3 11 (PE2), Improper duty (PC3)
Insufficient auditing of fishing vessel UI1-S 4.97E-3 12 Unsafe acts Improper actions for collision avoidance (AS1),
certificates Insufficient lookout (AJ3), Unsafe speeds (AS3), Failure to
Unreasonable navigation plan UP1 4.94E-3 13 detect other vessels (AS2), Misjudgment on collision risk
Unfamiliarity with collision PP2 4.45E-3 14 (AJ1)
avoidance rules
Insufficient experience and skills PP1 4.26E-3 15
3.4. Obtaining the development path

level to prevent collision accidents. The five levels of external factors, The strength of influence between any two adjacent nodes can be
organizational influences, unsafe supervision, preconditions for unsafe obtained using Eqs. (7)-(9). Table 12 shows the strength of influence,
acts and unsafe acts in HFACS are utilized as target nodes for sensitivity which can be used to discover the development path to the collision. The
analysis. Table 11 gives the sensitivity analysis results of these target “Average” value represents the average influence of two nodes. Addi­
nodes, with the bolded parts highlighting the special influencing factors tional details about other nodes are given in Appendix B. The develop­
of collisions between merchant ships and fishing vessels. ment path to the occurrence of the target node is identified by
Identifying the key influencing factors can be used to enhance connecting the adjacent nodes with the highest influence.
maritime safety in two aspects. For maritime authorities and fishing Taking “Collision” as an example, this process is illustrated as fol­
management departments, it can improve the shortcomings of existing lows. The factors with the highest influence on collisions are unsafe acts,
laws and regulations, strengthen weak points, and optimize the navi­ followed by preconditions for unsafe acts, unsafe supervision, and
gation environment. For merchant and fishing vessel operators, it will be organizational influences. The factors with the highest influence on
beneficial to identify the hazards conduct the risk assessment, and unsafe acts are skill-based errors, followed by judgment errors, pre­
introduce high technology, for example, intelligent collision avoidance conditions for unsafe acts, and violations. The factors with the highest
technology.

13
H. Wang et al. Reliability Engineering and System Safety 249 (2024) 110201

Table 12 Table 13
The strength of influence for some critical nodes. Some typical development paths of “collision”.
Parent nodes Child nodes Average Number The development paths

Organizational influences Collision 0.076 1 Improper actions for collision avoidance (AS1)→ Skill-based errors→
Preconditions for unsafe acts Collision 0.109 Unsafe acts→Collision
Unsafe acts Collision 0.148 2 Insufficient lookout (AJ3)→ Judgment errors→ Unsafe acts→ Collision
Unsafe supervision Collision 0.102 3 Crossing the shipping lane (AV3-S)→ Violations→ Unsafe acts→ Collision
Preconditions for unsafe acts Unsafe acts 0.127 4 PC1-S (Insufficient communication)→ Coordination factors→
Skill-based errors Unsafe acts 0.177 Preconditions for unsafe acts→ Unsafe acts→ Collision
Violations Unsafe acts 0.086 5 Unfamiliar with collision avoidance rules (PP2)→ Personnel factors→
Judgment errors Unsafe acts 0.128 Preconditions for unsafe acts→ Unsafe acts→Collision
Improper actions for collision avoidance (AS1) Skill-based errors 0.315 6 Insufficient auditing of fishing vessel certificates (UI1-S)→ Inadequate
Failure to detect other vessels (AS2) Skill-based errors 0.129 supervision→Unsafe supervision→ Preconditions for unsafe acts→ Unsafe
Unsafe speeds (AS3) Skill-based errors 0.199 acts→ Collision
7 Fishing methods (UF2-S)→ Failure to correct known problems→ Unsafe
supervision→ Preconditions for unsafe acts→ Unsafe acts→ Collision
influence on skill-based errors are improper actions for collision 8 Insufficient navigation safety education and training (OS1) → Safety
avoidance (AS1), followed by unsafe speeds (AS3), and failure to detect management→Organizational influences→ Unsafe supervision→
Preconditions for unsafe acts→Unsafe acts→ Collision
other vessels (AS2). By sequentially connecting these factors with the 9 Inadequate manning (OR1-S)→ Resource management→Organizational
highest influence, the development path of “Collision” is “Improper influences→ Unsafe supervision→ Preconditions for unsafe acts→ Unsafe
actions for collision avoidance (AS1)→ Skill-based errors→ Unsafe acts→ Collision
acts→ Collision”. Based on this method, the development paths of each 10 Insufficient regulation and enforcement (E1-S)→ External factors →
Organizational influences→ Unsafe supervision→ Preconditions for
node can be identified. Table 13 gives some typical development paths
unsafe acts→ Unsafe acts→ Collision
of “Collision”.

4. Discussion on the limitations of the study been identified. Moreover, the prominent development paths of collision
accidents are derived through the strength of influence analysis. Spe­
Although the proposed HFACS-BN model is useful for collision ac­ cifically, the development path is “Improper actions for collision
cident prevention by identifying the key contributing factors and avoidance (AS1) → Skill-based errors→ Unsafe acts→ Collision”.
development paths, some limitations should be discussed for the appli­ Although this paper focuses on the analysis of HOFs of collisions
cation of the proposed method. between merchant ships and fishing vessels in China, a similar situation
The HFACS-BN model is developed using data from collision acci­ in other countries could exist as fishing vessels are abundant in many
dents involving merchant ships and fishing vessels in China. Therefore, areas, such as the trawl fishing activities in the Northern and Central
when applying this model in other waters, it is necessary to make Adriatic Sea [85], fishing vessel activities in the Ascension Island [86],
appropriate adaptations based on the actual situation. For instance, the the activities of fishing vessels in Japan and Korea [87]. Therefore, the
influencing factors, including insufficient communication and inade­ modeling approach and findings in this paper can also be further applied
quate manning, are commonly recognized in China. However, when to the analysis of merchant and fishing vessel collisions in other areas. It
applying it to other waterways, this may be different and it should be is recommended that a complete dataset for specific areas should be
carefully handled because the HOFs may be slightly distinct in different collected to develop accident databases for collision accidents between
waterways of other countries, such as narrow channels, busy shipping merchant ships and fishing vessels in future work, which will be bene­
lanes, coastal areas, fishing areas, or the intersection area between the ficial to comprehensively identify the influencing factors. Moreover,
traditional routes of merchant ships and the traditional operating areas further analysis can be carried out to consider the influence of fishing
of fishing vessels. It can be verified more reasonably since the occur­ methods, which should be significant for preventing maritime accidents
rence likelihood of collision accidents between merchant ships and of fishing vessels.
fishing vessels can be calculated in specific waterways. However, as this Table 1,2,3,4,5,6,7,8,9,10,11,12,13,A1-B1
paper collects data from different waterways in China, this verification is
hard to be carried out and it will be improved in future studies. CRediT authorship contribution statement

5. Conclusions Hong Wang: Writing – original draft, Visualization, Methodology,

Investigation, Formal analysis, Data curation, Conceptualization. Ning
The main contribution of this paper is to propose an HFACS-BN Chen: Writing – original draft, Validation, Supervision, Methodology,
model to analyze the human and organizational factors and to drive Funding acquisition. Bing Wu: Writing – review & editing, Project
the development paths of collision accidents. Based on 443 accident administration, Methodology, Funding acquisition, Conceptualization.
data, 56 influencing factors including 11 special influencing factors for C. Guedes Soares: Writing – review & editing, Validation,
collision accidents between merchant ships and fishing vessels are Methodology.
identified using the modified HFACS framework, and they are trans­
formed into the BN for quantitative analysis of the HOFs. Compared to Declaration of competing interest
collisions between merchant ships, collisions between merchant ships
and fishing vessels are more influenced by organizational factors, unsafe The authors declare that they have no known competing financial
supervision, and preconditions for unsafe acts, but are slightly less interests or personal relationships that could have appeared to influence
influenced by unsafe acts. Sensitivity analysis is carried out to identify the work reported in this paper.
the key HOFs of collision accidents. The results indicate that unsafe acts
at the five levels of the HFACS are the predominant factors in the Data availability
occurrence of collision accidents, and insufficient regulation and
enforcement and improper actions for collision avoidance are key Data will be made available on request.
influencing factors among the top fifteen influencing factors of colli­
sions. The key factors that influence the five levels in HFACS have also

14
H. Wang et al. Reliability Engineering and System Safety 249 (2024) 110201

Acknowledgments 2023IVB079), Teaching Research Projects of Anhui Higher Education

Institutions (2021jyxm1335). The second author of this paper was
The research presented in this paper was sponsored by a grant from supported by the major science and technology program of Hainan
the National Natural Science Foundation of China (Grant No. Province (ZDKJ2020012), Yazhou Bay Science and Technology City
5191001041, 52071248 and 52272422), Innovation and Entrepre­ Supporting Project (SKJC-2022-PTDX-022), and Open Fund
neurship team import project of Shaoguan City (201212176230928), (2020KF0051).
Fundamental Research Funds for the Central Universities (WUT:

Appendix A. The typical keywords and sentences for identifying the influencing factors

Table A1
Some typical keywords and sentences.

Influencing factors Symbol Some typical keywords and sentences

Insufficient regulation and enforcement E1-S Inability to monitor vessels in the port approach channel via radar; insufficient management of fishing vessels operating at sea
with the required licenses; inadequate verification of crew staffing and certification status; the location of the accident is not
within the surveillance range of the VTS system; lack of safety monitoring of such off-site fishing vessels by fisheries authorities.
Inadequate information organization E2 Insufficient sharing of ship-to-shore information; AIS discrepancies and base station coverage issues sometimes result in vessels
not being displayed on the regulatory platform even when they are operational; insufficient informatization in fishery port
supervision; no information on accidental fishing vessels found on the Fishing Port Pass platform; the fishing vessel is not
equipped with the Beidou system for fisheries, etc.
Imperfect management mechanisms E3-S Lack of focused joint supervision by maritime and fisheries authorities in areas where the customary routes of merchant ships
intersect with fishing; insufficient joint maritime patrols for law enforcement; insufficient joint publicity and education; the
division of shipping routes for merchant ships and fishing areas is unclear, with irrational boundaries, thus highlighting the
prominent contradiction in sea usage between merchant ships and fishing vessels.
Inadequate manning OR1-S Lack of crew certification; crew certification is not of the specified grade; non-compliance with the vessel’s Minimum Safe
Manning Certificate requirements; the crews are insufficiently equipped and qualified to meet the requirements; fishermen
lacking a valid crew certificate.
Emergency management OR2 Insufficient emergency drills; inadequate life-saving equipment; life jackets and lifebuoys not equipped with self-igniting lights;
insufficient emergency response capability of crew members in emergencies, etc.
Insufficient facilities and equipment OR3 Lack of radar equipment; failure to install the Automatic Identification System; failure to install Very High Frequency
equipment; lack of navigation, signaling, and communication devices; failure to install navigation lights as regulations; the
corrosion of the masthead and stern lights has resulted in their inability to function properly; the presence of a ship’s self-
equipped AIS device.
Lack of effective administration of OO1-S Lax management of crew duty on individually operated fishing vessels; lack of safety management system on individually
fishing vessels operated fishing vessels; fishermen fishing on vessels without professional training, etc.
Failure to implement regulations or OO2 Violation of “the Regulations on the Watchkeeping of Seafarers of the People’s Republic of China,” violation of the “Inland
rules Water Crew Watchkeeping Regulations of the People’s Republic of China,” violation of STCW Rules, violation of the “Fishing
Vessel Navigation Watchkeeping Guidelines,” violation of the “Maritime Traffic Safety Law of the People’s Republic of China,”
violation of the “Fisheries Catch Permit Management Regulations,” violation of the “Regulations on the Management of Vessel
Pilotage,” violation of the “Fishery Law of the People’s Republic of China,” etc.
Communication barriers OO3-S Different VHF radio communication frequencies; weak radar reflectivity of wooden fishing vessels; high radar clutter;
fishermen do not speak Mandarin, only the local dialect; fishermen are unable to correctly understand instructions; adverse
external communication environments, such as noise, strong winds, and waves.
Insufficient navigation safety education OS1 Insufficient safety education; insufficient training in emergency ship handling skills and correct use of navigation instruments;
and training inadequate warning education through typical cases of merchant and fishing vessel collisions; failure to provide practical safety
training; insufficient training in the timing and operation of avoiding maneuvers when merchant ships and fishing vessels
encounter each other.
Poor safety culture OS2 The work done by the shipping company in creating a safety culture and raising the safety awareness of its employees is
inadequate; the inadequacy of the safety management system; insufficient safety awareness among crew members, etc.
Insufficient safety assessment OS3 Failure of shipping companies to equip their ships with effective early warning systems; fishing vessels lack the technology and
technology systems necessary for tracking the movements of other vessels; insufficient capacity of fishermen to accurately assess the
situation and risks around them, etc.
Insufficient auditing of fishing vessel UI1-S Fishing vessels without a valid vessel inspection certificate, vessel registration certificate, or fishing license; “three No’s” fishing
certificates vessels; unlicensed small fishing vessels; insufficient fishing vessel certificates.
Inadequate assessment of education and UI2 Crew incompetency; insufficient collision avoidance skills; issues of unexamined personnel and certificate mismatches, etc.
training
Personal factors of the supervisor UI3 Failure of the captain to fulfill his duties; unexplained switching off AIS; the captain is not in the ship’s cockpit.
Insufficient risk assessment UF1 No recognition of an imminent dangerous trend; entering a critical situation without recognizing its danger; navigating in
heavy fog conditions; failure to identify the potential risks of complex navigation environments, etc.
Fishing methods UF2-S Fishing vessels are conducting fishing operations without paying attention to vessels traveling in the vicinity; merchant ships
have not found the nets of fishing vessels trawling for fish; improper placement of net position instruments.
Insufficient safety reporting UF3 Failure to report the ship’s entry or exit from port; failure to notify the captain upon detecting danger; failure to inform other
crew members upon detecting danger.
Unreasonable navigation plan UP1 Navigating on routes unsuitable for merchant ships; venturing to navigate under conditions that do not meet the requirements
for nighttime navigation; navigating beyond designated navigation areas and routes, and violating lane navigation systems;
choosing to navigate when the visibility distance is significantly less than the safe navigation distance and failing to anchor
promptly when possible; venturing to navigate without meeting the conditions for safe navigation.
Insufficient supervision plan UP2 Lack of self-supervisory organizations for fishing vessels; lack of self-supervisory systems for fishing vessels; failure of the
shipping company to fulfill its supervisory responsibilities; lack of a supervision plan, etc.
Unaware of regulations and legislation UP3-S Low awareness of laws and regulations among fishermen; engaging in fishing operations in violation of regulations.
(continued on next page)

15
H. Wang et al. Reliability Engineering and System Safety 249 (2024) 110201

Table A1 (continued )
Influencing factors Symbol Some typical keywords and sentences

Insufficient experience and skills PP1 Insufficient crew experience; insufficient crew skills; lack of familiarity with the operational and navigational characteristics of
fishing vessels, etc.
Unfamiliarity with collision avoidance PP2 Crew members need to strengthen training on collision avoidance rules; failure to understand the “People’s Republic of China
rules Inland River Collision Avoidance Regulations”; failure to understand the COLREGs.
Physical or mental limitations PP3 Navigating for a long duration without rest; drinking alcohol; fatigued driving; poor physical condition; engaging in off-task
activities; listening to music while on duty, etc.
Insufficient communication PC1-S No response to calls; failure to establish contact with “X” ship; no effective communication.
Inadequate use of sound or light signals PC2 Failure to display navigation lights as required; failure to display fog lights as required; fishing vessels do not display trawling
operation lights as required; failure to display anchor signal lights as required; failure to switch on navigational lights; failure to
sound signals as required.
Improper duty PC3 Failure to fulfill duty obligations; distracted with other matters while on duty; failure to increase the number of the crews on
duty in poor visibility; failure to increase the number of the crews on duty in heavy traffic water areas; watchkeeping personnel
positions do not meet the requirements; insufficient number of the crews on duty; failure to arrange for crews to be on duty.
Restricted visibility PE1 Poor visibility; restricted visibility.
Complex navigational environment PE2 Complex navigational environment; dense fishing nets; heavy traffic flow; intersecting areas; narrow channels; the traffic flow
of fishing vessels crossing this area does not have a fixed direction, leading to a complex and diverse situation of merchant ships
encountering each other in this area; the encounter situation between merchant ships and fishing vessels is complicated.
Meteorological and hydrological PE3 Strong winds, big waves, fast water flow, heavy fog, etc.
conditions
Misjudgment on collision risk AJ1 Inadequate judgment of collision risk; failure to make adequate estimates of collision risk.
Incorrect use of navigational aids AJ2 Improper radar range settings; failure to perform radar plotting; non-use of VHF; failure to detect the other ship’s AIS signal.
Insufficient lookout AJ3 Neglect of lookout; failure to maintain a regular lookout; failure to increase the number of lookout personnel; failure to
maintain a regular lookout by visual, auditory, and all other effective means.
Improper actions for collision avoidance AS1 Improper actions and measures to avoid a collision; failure to take actions most favorable to collision avoidance; no collision
avoidance actions.
Failure to detect other vessels AS2 Failure to detect the other ship; failure to detect the other vessel in time; both sides still not detecting each other; failure to
detect the operating fishing vessel; failure to detect the merchant ship.
Unsafe speeds AS3 Failure to use safe speed; failure to reduce the ship’s speed in time; taking measures to slow down only when a collision is
imminent; ships navigating too fast.
Illegal fishing AV1-S Fishing within port navigation channels; fishing at anchorages; fishing during closed seasons; cross-boundary fishing
operations; fishing by unlicensed, unregistered, and uninspected fishing vessels.
Illegal navigation of merchant ships AV2 Overloading of merchant ships; traversing beyond designated shipping routes; navigating within restricted areas; illegal
construction activities during navigation, etc.
Crossing the shipping lane AV3-S Improper crossing of the shipping lane; crossing the shipping lane in violation of the regulations.

Appendix B. The strength of influence for typical nodes

Table B1
The strength of influence for typical nodes.

Parent nodes Child nodes Average

Illegal fishing (AV1-S) Violations 0.192

Illegal navigation of merchant ships (AV2) Violations 0.182
Crossing the shipping lane (AV3-S) Violations 0.235
Inappropriate operations Unsafe supervision 0.159
Failure to correct known problems Unsafe supervision 0.151
Inadequate supervision Unsafe supervision 0.143
Organizational influences Unsafe supervision 0.185
Insufficient navigation safety education and training (OS1) Safety management 0.387
Poor safety culture (OS2) Safety management 0.149
Insufficient safety assessment technology (OS3) Safety management 0.226
Inadequate manning (OR1-S) Resource management 0.300
Emergency management (OR2) Resource management 0.147
Insufficient facilities and equipment (OR3) Resource management 0.199
Coordination factors Preconditions for unsafe acts 0.225
Environmental factors Preconditions for unsafe acts 0.137
Personnel factors Preconditions for unsafe acts 0.258
Unsafe supervision Preconditions for unsafe acts 0.140
Insufficient experience and skills (PP1) Personnel factors 0.264
Unfamiliarity with collision avoidance rules (PP2) Personnel factors 0.317
Physical or mental limitations (PP3) Personnel factors 0.123
External factors Organizational influences 0.190
Organizational climate and process Organizational influences 0.085
Resource management Organizational influences 0.129
Safety management Organizational influences 0.149
Lack of effective administration of fishing vessels (OO1-S) Organizational climate and process 0.191
Failure to implement regulations or rules (OO2) Organizational climate and process 0.245
Communication barriers (OO3-S) Organizational climate and process 0.152
Misjudgment on collision risk (AJ1) Judgment errors 0.180
Incorrect use of navigational aids (AJ2) Judgment errors 0.103
Insufficient lookout (AJ3) Judgment errors 0.303
(continued on next page)

16
H. Wang et al. Reliability Engineering and System Safety 249 (2024) 110201

Table B1 (continued )
Parent nodes Child nodes Average

Insufficient auditing of fishing vessel certificates (UI1-S) Inadequate supervision 0.275

Inadequate assessment of education and training (UI2) Inadequate supervision 0.253
Personal factors of the supervisor (UI3) Inadequate supervision 0.185
Insufficient risk assessment (UF1) Failure to correct known problems 0.265
Fishing methods (UF2-S) Failure to correct known problems 0.264
Insufficient safety reporting (UF3) Failure to correct known problems 0.166
Insufficient regulation and enforcement (E1-S) External factors 0.248
Inadequate information organization (E2) External factors 0.192
Imperfect management mechanisms (E3-S) External factors 0.170
Restricted visibility (PE1) Environmental factors 0.216
Complex navigational environment (PE2) Environmental factors 0.274
Meteorological and hydrological conditions (PE3) Environmental factors 0.164
Insufficient communication (PC1-S) Coordination factors 0.314
Inadequate use of sound or light signals (PC2) Coordination factors 0.145
Improper duty (PC3) Coordination factors 0.238
Unreasonable navigation plan (UP1) Inappropriate operations 0.256
Insufficient supervision plan (UP2) Inappropriate operations 0.266
Unaware of regulations and legislation (UP3-S) Inappropriate operations 0.182

References [21] Chen P, Zhang Z, Huang Y, Dai L, Hu H. Risk assessment of marine accidents with
Fuzzy Bayesian Networks and causal analysis. Ocean Coast Manag 2022;228:
106323.
[1] Zhang J, Teixeira ÂP, Guedes Soares C, Yan X. Quantitative assessment of collision
[22] Zhao C, Yip TL, Wu B, Lyu J. Use of fuzzy fault tree analysis and Bayesian network
risk influence factors in the Tianjin port. Saf Sci 2018;110:363–71.
for occurrence likelihood estimation of navigational accidents in the Qinzhou Port.
[2] Cheng Z, Zhang Y, Wu B, Guedes Soares C. Traffic-conflict and fuzzy-logic-based
Ocean Eng 2022;263:112381.
collision risk assessment for constrained crossing scenarios of a ship. Ocean Eng
[23] Antão P, Sun S, Teixeira AP, Guedes Soares C. Quantitative assessment of ship
2023;274:114004.
collision risk influencing factors from worldwide accident and fleet data. Reliab
[3] Yu Y, Chen L, Shu Y, Zhu W. Evaluation model and management strategy for
Eng Syst Saf 2023;234:109166.
reducing pollution caused by ship collision in coastal waters. Ocean Coast Manag
[24] Li Z, Hu S, Zhu X, Gao G, Yao C, Han B. Using DBN and evidence-based reasoning
2021;203:105446.
to develop a risk performance model to interfere ship navigation process safety in
[4] Mazurek J, Lu L, Krata P, Montewka J, Krata H, Kujala P. An updated method
Arctic waters. Process Safety Environ Protect 2022;162:357–72.
identifying collision-prone locations for ships. A case study for oil tankers
[25] Fu S, Yu Y, Chen J, Xi Y, Zhang M. A framework for quantitative analysis of the
navigating in the Gulf of Finland. Reliab Eng Syst Saf 2022;217:108024.
causation of grounding accidents in arctic shipping. Reliab Eng Syst Saf 2022;226:
[5] Cai M, Zhang J, Zhang D, Yuan X, Guedes Soares C. Collision risk analysis on ferry
108706.
ships in Jiangsu Section of the Yangtze River based on AIS data. Reliab Eng Syst Saf
[26] Li X, Chen C, Yang F-q. Exploring hazardous chemical explosion accidents with
2021;215:107901.
association rules and Bayesian networks. Reliab Eng Syst Saf 2023;233:109099.
[6] Yu Q, Liu K, Yang Z, Wang H, Yang Z. Geometrical risk evaluation of the collisions
[27] Zhang M, Zhang D, Goerlandt F, Yan X, Kujala P. Use of HFACS and fault tree
between ships and offshore installations using rule-based Bayesian reasoning.
model for collision risk factors analysis of icebreaker assistance in ice-covered
Reliab Eng Syst Saf 2021;210:107474.
waters. Saf Sci 2019;111:128–43.
[7] Korçak M, Balas CE. Reducing the probability for the collision of ships by changing
[28] Ugurlu O, Yildiz S, Loughney S, Wang J, Kuntchulia S, Sharabidze I. Analyzing
the passage schedule in Istanbul Strait. Int J Disaster Risk Red 2020;48:101593.
collision, grounding, and sinking accidents occurring in the black sea utilizing
[8] Wang M, Wang Y, Cui E, Fu X. A novel multi-ship collision probability estimation
HFACS and bayesian networks. Risk Anal 2020;40:2610–38.
method considering data-driven quantification of trajectory uncertainty. Ocean
[29] Qiao W, Liu Y, Ma X, Liu Y. A methodology to evaluate human factors contributed
Eng 2023;272:113825.
to maritime accident by mapping fuzzy FT into ANN based on HFACS. Ocean Eng
[9] Xu S, Kim E. Hybrid causal logic model for estimating the probability of an
2020;197:106892.
icebreaker–ship collision in an ice channel during an escort operation along the
[30] Yildiz S, Uğurlu Ö, Wang J, Loughney S. Application of the HFACS-PV approach for
Northeast Passage. Ocean Eng 2023;284:115264.
identification of human and organizational factors (HOFs) influencing marine
[10] Lan H, Ma X, Qiao W, Deng W. Determining the critical risk factors for predicting
accidents. Reliab Eng Syst Saf 2021;208:107395.
the severity of ship collision accidents using a data-driven approach. Reliab Eng
[31] Ma L, Ma X, Lan H, Liu Y, Deng W. A data-driven method for modeling human
Syst Saf 2023;230:108934.
factors in maritime accidents by integrating DEMATEL and FCM based on HFACS:
[11] Wu B, Tian H, Yan X, Guedes Soares C. A probabilistic consequence estimation
a case of ship collisions. Ocean Eng 2022;266:112699.
model for collision accidents in the downstream of Yangtze River using Bayesian
[32] Reason J. Human error. New York: Cambridge university press; 1990.
Networks. Proc Inst Mech Eng, Part O: J Risk Reliab 2019;234:422–36.
[33] Shappell SA, Wiegmann DA. The human factors analysis and classification
[12] Domeh V, Obeng F, Khan F, Bose N, Sanli E. An operational risk awareness tool for
system–hfacs.(dot/faa/am-00/7). 2000. Retrieved May 15, 2023 from, https://ros
small fishing vessels operating in harsh environment. Reliab Eng Syst Saf 2023;
ap.ntl.bts.gov/view/dot/15397.
234:109139.
[34] Wu B, Yip TL, Yan X, Guedes Soares C. Review of techniques and challenges of
[13] Wang H, Liu Z, Wang X, Graham T, Wang J. An analysis of factors affecting the
human and organizational factors analysis in maritime transportation. Reliab Eng
severity of marine accidents. Reliab Eng Syst Saf 2021;210:107513.
Syst Saf 2022;219:108249.
[14] FAO. Fishing safety. Food and Agriculture Organization of the United Nations; (n.
[35] Wiegmann DA, Shappell SA. A human error approach to aviation accident analysis:
d). Retrieved May 15, 2023 from https://www.fao.org/fishing-safety/risk-manage
the human factors analysis and classification system: reference & research book.
ment/occupational-hazards/zh/.
News; 2003.
[15] Kaptan M, Sarialioglu S, Ugurlu O, Wang J. The evolution of the HFACS method
[36] Yıldırım U, Başar E, Uğurlu Ö. Assessment of collisions and grounding accidents
used in analysis of marine accidents: a review. Int J Ind Ergon 2021;86:103225.
with human factors analysis and classification system (HFACS) and statistical
[16] Antao P, Guedes Soares C. Analysis of the influence of human errors on the
methods. Saf Sci 2019;119:412–25.
occurrence of coastal ship accidents in different wave conditions using Bayesian
[37] Shi X, Zhuang H, Xu D. Structured survey of human factor-related maritime
Belief Networks. Accident Anal Prevent 2019;133:105262.
accident research. Ocean Eng 2021;237:109561.
[17] Sotiralis P, Ventikos NP, Hamann R, Golyshev P, Teixeira A. Incorporation of
[38] McGuinness E, Utne IB. A systems engineering approach to implementation of
human factors into ship collision risk models focusing on human centred design
safety management systems in the Norwegian fishing fleet. Reliab Eng Syst Saf
aspects. Reliab Eng Syst Saf 2016;156:210–27.
2014;121:221–39.
[18] Aydin M, Akyuz E, Turan O, Arslan O. Validation of risk analysis for ship collision
[39] Celik M, Cebi S. Analytical HFACS for investigating human errors in shipping
in narrow waters by using fuzzy Bayesian networks approach. Ocean Eng 2021;
accidents. Accid Anal Prev 2009;41:66–75.
231:108973.
[40] Li Z, Wang X, Gong S, Sun N, Tong R. Risk assessment of unsafe behavior in
[19] Fan S, Blanco-Davis E, Yang Z, Zhang J, Yan X. Incorporation of human factors into
university laboratories using the HFACS-UL and a fuzzy Bayesian network. J Safety
maritime accident analysis using a data-driven Bayesian network. Reliab Eng Syst
Res 2022;82:13–27.
Saf 2020;203:107070.
[41] MSA. Accident investigation reports. Maritime Safety Administration of the
[20] Coraddu A, Oneto L, Navas de Maya B, Kurt R. Determining the most influential
People’s Republic of China; 2013-2023. Retrieved February 18, 2024 from, htt
human factors in maritime accidents: a data-driven approach. Ocean Eng 2020;
ps://www.msa.gov.cn/page/outter/shiguyufang.jsp.
211:107588.

17
H. Wang et al. Reliability Engineering and System Safety 249 (2024) 110201

[42] SMSA. Accident investigation reports. Shandong Maritime Safety Administration; [65] Zhang X, Chen W, Xi Y, Hu S, Tang L. Dynamics Simulation of the Risk Coupling
2013-2023. Retrieved February 18, 2024 from, https://www.sd.msa.gov.cn/col/ Effect between Maritime Pilotage Human Factors under the HFACS Framework.
col15701/index.html. J Mar Sci Eng 2020;8:144.
[43] HMSA. Maritime safety accidents. Hebei Maritime Safety Administration; 2014- [66] Yang J, Sun Y, Song Q, Ma L. Laws and preventive methods of collision accidents
2023. Retrieved February 18, 2024 from, https://www.hb.msa.gov.cn/htm between merchant and fishing vessels in coastal area of China. Ocean Coast Manag
l/ssaqsg/. 2023;231:106404.
[44] LMSA. Safety and emergency information. Liaoning Maritime Safety [67] Rezaee S, Pelot R, Ghasemi A. The effect of extreme weather conditions on
Administration; 2017-2023. Retrieved February 18, 2024 from, https://www.ln. commercial fishing activities and vessel incidents in Atlantic Canada. Ocean Coast
msa.gov.cn/xxgk/xxgkml/ssjtsgxx/index_3.shtml. Manag 2016;130:115–27.
[45] HMSA. Safety supervision. Hainan Maritime Safety Administration; 2020-2023. [68] Jin D. The determinants of fishing vessel accident severity. Accident Anal Prevent
Retrieved February 18, 2024 from, https://www.hn.msa.gov.cn/hsfw_9/index.jh 2014;66:1–7.
tml. [69] Uğurlu F, Yıldız S, Boran M, Uğurlu Ö, Wang J. Analysis of fishing vessel accidents
[46] LMSA. Information on maritime safety accidents. Lianyungang Maritime Safety with Bayesian network and Chi-square methods. Ocean Eng 2020;198:106956.
Administration; 2017-2023. Retrieved February 18, 2024 from, https://www.lyg. [70] Wang S, Zhang S, Tang F, Shi Y, Sui Y, Fan X, et al. Developing machine learning
msa.gov.cn/ssaqsgxx/index.jhtml?lKuyjUEbpua1=1711339750317#mainContain methods for automatic recognition of fishing vessel behaviour in the Scomber
er. japonicus fisheries. Front Mar Sci 2023;10.
[47] TMSA. Maritime investigations. Tianjin Maritime Safety Administration; 2019- [71] Cheng T, Utne IB, Wu B, Wu Q. A novel system-theoretic approach for human-
2023. Retrieved February 18, 2024 from, https://www.tj.msa.gov.cn/gov-public. system collaboration safety: case studies on two degrees of autonomy for
html?id=2. autonomous ships. Reliab Eng Syst Saf 2023;237:109388.
[48] SMSA. Maritime investigations. Shanghai Maritime Safety Administration; 2015- [72] Xia N, Zou PXW, Liu X, Wang X, Zhu R. A hybrid BN-HFACS model for predicting
2023. Retrieved February 18, 2024 from, https://www.sh.msa.gov.cn/xxgk/index. safety performance in construction projects. Saf Sci 2018;101:332–43.
jhtml?v=865. [73] Ghasemi F, Gholamizadeh K, Farjadnia A, Sedighizadeh A, Kalatpour O. Human
[49] FMSA. Maritime investigations. Fujian Maritime Safety Administration; 2019- and organizational failures analysis in process industries using FBN-HFACS model:
2023. Retrieved February 18, 2024 from, https://www.fj.msa.gov.cn/fjmsacm learning from a toxic gas leakage accident. J Loss Prev Process Ind 2022;78.
s/cms/html/fjhsjwwwz/index/1150764331.html. [74] Qiao W, Liu Y, Ma X, Liu Y. Human factors analysis for maritime accidents based
[50] GMSA. Information on maritime safety accidents. Guangxi Maritime Safety on a dynamic fuzzy bayesian network. Risk Anal 2020;40:957–80.
Administration; 2015-2023. Retrieved February 18, 2024 from, http://www.gx. [75] Fu L, Wang X, Wang D, Griffin MA, Li P. Human and organizational factors within
msa.gov.cn/contentlist.html?classid=173. the public sectors for the prevention and control of epidemic. Saf Sci 2020;131:
[51] ZMSA. Maritime safety accidents. Zhejiang Maritime Safety Administration; 2016- 104929.
2023. Retrieved February 18, 2024 from, https://www.zj.msa.gov.cn/ZJ/zwgk/ [76] Zhang J, Jin M, Wan C, Dong Z, Wu X. A Bayesian network-based model for risk
gkml/xzqz/index_1.html. modeling and scenario deduction of collision accidents of inland intelligent ships.
[52] Lan H, Ma X, Qiao W, Ma L. On the causation of seafarers’ unsafe acts using Reliab Eng Syst Saf 2023:109816.
grounded theory and association rule. Reliab Eng Syst Saf 2022;223:108498. [77] Ji Z, Xia Q, Meng G. A review of parameter learning methods in Bayesian network.
[53] Dirhamsyah D, Umam S, Arifin Z. Maritime law enforcement: indonesia’s In: Advanced Intelligent Computing Theories and Applications: 11th International
experience against illegal fishing. Ocean Coast Manag 2022;229:106304. Conference, ICIC 2015. Fuzhou, China: Springer; 2015. p. 3–12. August 20-23,
[54] Lazakis I, Kurt RE, Turan O. Contribution of human factors to fishing vessel 2015 Proceedings, Part III 11.
accidents and near misses in the UK. J Shipping Ocean Eng 2014;4:245–61. [78] Mustafaa Y, Tolpekin V, Stein A. Application of the EM-algorithm for Bayesian
[55] Obeng F, Domeh V, Khan F, Bose N, Sanli E. Capsizing accident scenario model for network modelling to improve forest growth estimates. Procedia Environ Sci 2011;
small fishing trawler. Saf Sci 2022;145:105500. 7:74–9.
[56] Obeng F, Domeh V, Khan F, Bose N, Sanli E. Analyzing operational risk for small [79] Wu B, Tang Y, Yan X, Guedes Soares C. Bayesian Network modelling for safety
fishing vessels considering crew effectiveness. Ocean Eng 2022;249:110512. management of electric vehicles transported in RoPax ships. Reliab Eng Syst Saf
[57] Özaydın E, Fışkın R, Uğurlu Ö, Wang J. A hybrid model for marine accident 2021:209.
analysis based on Bayesian Network (BN) and Association Rule Mining (ARM). [80] MemberIEE Castillo E, Ge JeM, Hadi AS. Sensitivity analysis in discrete bayesian
Ocean Eng 2022;247:110705. networks. IEEE Trans Syst, Man, Cybernet - Part A: Syst Humans 1997;27:412–23.
[58] Kincl L, Syron L, Lucas D, Vaughan A, Bovbjerg V. Relationship of personal, [81] Kjærulff CUL, vd Gaag. Making sensitivity analysis computationally efficient. In:
situational, and environmental factors to injury experience in commercial fishing. Proceedings of the Sixteenth Annual Conference on Uncertainty in Artificial
J Safety Res 2023;87:375–81. Intelligence; 2000. p. 317–25.
[59] Jiang D, Wu B, Cheng Z, Xue J, van Gelder PHAJM. Towards a probabilistic model [82] Koiter JR. Visualizing inference in Bayesian networks. Delft University of
for estimation of grounding accidents in fluctuating backwater zone of the Three Technology; 2006. p. 855.
Gorges Reservoir. Reliab Eng Syst Saf 2021;205:107239. [83] Hellinger E. Die orthogonalinvarianten quadratischer formen von unendlichvielen
[60] Davis B, Colbourne B, Molyneux D. Analysis of fishing vessel capsizing causes and variabelen. W. Fr. Kaestner; 1907.
links to operator stability training. Saf Sci 2019;118:355–63. [84] Adumene S, Afenyo M, Salehi V, William P. An adaptive model for human factors
[61] Oh J-H, Kim K-i, Jeong J-S. A study on the risk analysis based on the trajectory of assessment in maritime operations. Int J Ind Ergon 2022;89:103293.
fishing vessels in the VTS area. Int J e-Navigat Maritime Econ 2015;2:38–46. [85] Russo E, Monti MA, Mangano MC, Raffaetà A, Sarà G, Silvestri C, et al. Temporal
[62] Míguez González M, Bulian G. Influence of ship dynamics modelling on the and spatial patterns of trawl fishing activities in the Adriatic Sea (Central
prediction of fishing vessels roll response in beam and longitudinal waves. Ocean Mediterranean Sea, GSA17). Ocean Coast Manag 2020;192:105231.
Engineering 2018;148:312–30. [86] Rowlands G, Brown J, Soule B, Boluda PT, Rogers AD. Satellite surveillance of
[63] Fu S, Zhang Y, Zhang M, Han B, Wu Z. An object-oriented Bayesian network model fishing vessel activity in the ascension island exclusive economic zone and marine
for the quantitative risk assessment of navigational accidents in ice-covered Arctic protected area. Mar Policy 2019;101:39–50.
waters. Reliab Eng Syst Saf 2023;238:109459. [87] Sun Y, Lian F, Yang Z. Analysis of the activities of high sea fishing vessels from
[64] Li H, Ren X, Yang Z. Data-driven Bayesian network for risk analysis of global China, Japan, and Korea via AIS data mining. Ocean Coast Manag 2023;242:
maritime accidents. Reliab Eng Syst Saf 2023;230:108938. 106690.

18