3
1 INTRODUCTION
Maritime transport plays a vital role in international
trade and the global economy, with 70% of hazardous
goods being transported by sea. As demand for
transport increases, maritime safety management faces
greater challenges, and the systematic identification
and analysis of key RIFs is crucial for ensuring
navigation safety and achieving sustainable
development.
Numerous studies have been conducted on
maritime risk management of hazardous cargo
transport. For example, Elidolu et al. (2024) assessed
the risks associated with the transportation of styrene
monomer, identifying 24 failure modes using the
FMECA and fuzzy logic methods and proposing
preventive measures, thereby enhancing risk
awareness in the operation of liquid chemical ships. Ma
et al. (2023) investigated the coupled links in the
maritime transportation system of hazardous
chemicals, finding that the risks in the links of
consignment and transportation, as well as loading and
navigation, were relatively high. By quantifying risks
through fuzzy set theory and risk matrix, they
provided a basis for optimizing risk management and
emergency response strategies. These methods offer
references for the safety management of hazardous
cargo vessels in maritime transportation, but they have
certain limitations in dealing with complexity and
uncertainty.
In recent years, Bayesian networks (BN) have
become a new hotspot in maritime risk assessment due
to their advantages in describing the relationships
between random variables and risk reasoning. Khan et
Analysing Risk Influencing Factors for the Navigational
Safety of Hazardous Cargo Vessels Using Bayesian
Networks
C.P. Wan
1
, J.Z. Qi
1
, X.B. Wang
3
, Z.S. Yang
2
, B. Wu
1
, D. Zhang
1
& X.P. Yan
1
1
Wuhan University of Technology, Wuhan, China
2
Shenzhen Technology University, Shenzhen, China
3
COSCO Shipping Technology Co., Ltd, Shanghai, China
ABSTRACT: Enhancing navigational safety of hazardous cargo vessels constitutes a critical imperative for
sustaining maritime transportation system stability and fostering sustainable industry development. Based on
the developed database containing 106 accident reports involving hazardous cargo vessels collected from the
International Maritime Organization (IMO), this study aims to analyze the key risk influencing factors (RIFs)
contributing to the maritime traffic accidents. Utilizing text analysis, the research first identifies critical RIFs
across five primary domains, which are human, vessel, cargo, environment, and management. A Bayesian
network model is subsequently developed to map out the interrelationships among these identified navigational
safety RIFs. The findings suggest that factors such as "insufficient personnel training," "inadequate safety
inspections," "flammable and explosive cargo," "inadequate hazardous goods management," and "pollutant and
toxic cargo" exert the most pronounced influence on maritime traffic accidents. Based on these pivotal RIFs and
their evolutionary trajectories, this paper can offer theoretical support for enhancing the navigational safety of
hazardous cargo vessels.
http://www.transnav.eu
the International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 20
Number 1
March 2026
DOI: 10.12716/1001.20.01.01
4
al. (2024) evaluated the risks of hazardous chemicals in
seaports using BN, identifying fire, corrosion, and
improper handling as the main RIFs and proposing
measures to reduce risks. Cao et al. (2023) analyzed the
relationship between the severity of maritime accidents
and related factors using a data-driven BN model,
finding that accident type, main engine power, gross
tonnage, ship type, and accident location were key
influencing factors, thus enabling the prediction of
severe accidents. Sezer and Akyuz (2024) combined the
evidence reasoning - success possibility index method
with Bayesian belief networks to construct a risk model
for fire/explosion in the cargo hold of a chemical ship,
identifying the main risks in the cleaning operation.
Sezer et al. (2023) conducted a quantitative analysis of
the oil spill risk during the crude oil loading process of
an oil tanker using an improved Z-number theory and
BN, assessing the probability of oil spill accidents and
supporting risk decision-making and accident
prevention. Wu et al. (2021) proposed a Liquefied
Natural Gas (LNG) risk assessment model based on the
Bayesian - mutation theory - energy transfer method,
identifying the key RIFs in port LNG bunkering and
storage, and enhancing the capability of safety
prevention and control. Fan et al. (2020) used BN and
TOPSIS technology to formulate maritime accident
prevention strategies from a human factors
perspective, emphasizing the importance of
information communication, command system, and
safety culture, and improving the effectiveness of
accident prevention. Chen et al. (2022) assessed the
risks of maritime transportation based on fuzzy BN
and causal analysis, identifying and quantifying the
main RIFs such as attention distraction and adverse
weather. Shi (2017) analyzed the dynamic risks of
hazardous chemicals shipping from personnel, ship,
environment, and cargo aspects. Guo et al. (2023)
analyzed the risk coupling of hazardous materials
transportation accidents in complex maritime
environments, categorizing risks into six types and
identifying 22 factors, and developed an AHP-N-K
model to assess the coupling degree and its correlation
mechanisms. Khan et al. (2021) conducted a study to
assess the risks associated with the berthing of
hazardous cargo vessels, identifying key RIFs and their
influence on maritime accidents, with a focus on
human and environmental elements as the most
significant contributors.
Therefore, in this study, a data-driven BN model is
proposed to identify key RIFs affecting hazardous
cargo vessels navigation, as well as assess their exact
influence on maritime traffic accidents. This study
initially organized relevant accident reports to identify
primary RIFs from five aspects: personnel, vessel,
cargo, environment, and management. Subsequently, a
BN model for the risk elements of hazardous cargo
vessels navigation safety was constructed based on the
identified RIFs and their interrelationships. Through
sensitivity analysis and node strength analysis, key
RIFs and their evolution paths were identified, and
corresponding control measures were proposed. These
research findings aim to provide theoretical support
for ensuring the navigation safety of hazardous cargo
vessels.
The paper is structured as follows: Section 2
introduces the classification of hazardous cargo
vessels. Section 3 outlines the research methods.
Section 4 presents the risk factor framework. Section 5
constructs a navigation risk model using BN. Section 6
identifies critical RIFs and their trajectories through
sensitivity and node strength analyses. Finally, Section
7 summarizes the conclusions.
2 CLASSIFICATION OF HAZARDOUS CARGO
VESSELS
In accordance with the International Maritime
Dangerous Goods Code (IMDG Code)(2025),
dangerous chemicals are substances that, due to their
chemical or physical properties, may pose a risk to
human health, safety, property, or the environment
during transportation, loading/unloading, and storage.
Therefore, a chemical tanker is defined as a vessel
specifically designed for transporting such dangerous
chemicals, and its design, construction, and operation
must comply with the requirements of the IMDG Code,
including the classification, packaging, marking,
stowage, and segregation of the cargo, to ensure the
safety of the transportation process and environmental
protection.
Based on relevant literature and data, hazardous
cargo vessels are divided into four categories: liquid
chemical tankers, oil tankers (oil ships), liquefied gas
carriers, and chemical bulk carriers, as shown in
Table 1.
Table 1. Hazardous cargo vessels classification and features
comparison table
Classification
Common
Cargo
Types
Cargo Hold
Structure Features
Safety
Requirements
and Regulations
C1 Oil
Tankers
Crude
oil,
diesel,
gasoline,
heavy oil
Compartmentalized
cargo holds;
complete ballast
system
Complies with
the International
Bulk Chemical
Code (IBC Code),
categorized into
Types I, II, and III
based on the
hazard level of
the cargo
C2 Liquid
Chemical
Tankers
Petroleu
m
chemical
s,
vegetable
oils,
organic
solvents,
etc.
Double bottoms;
corrosion-resistant
materials
Complies with
the International
Bulk Chemical
Code (IBC Code),
categorized into
Types I, II, and III
based on the
hazard level of
the cargo
C3 Liquefied
Gas Carriers
LNG,
Liquefie
d
Petroleu
m Gas
(LPG),
ammonia
, etc.
Complex structure;
equipped with
temperature and
pressure control
systems
Complies with
IBC Code and
relevant gas
transport
regulations, such
as specific
requirements for
LNG and LPG
C4 Chemical
Bulk Carriers
Soap,
caustic
soda,
fertilizers
, etc.
Specialized cargo
holds; moisture-
proof and anti-
pollution
Complies with
the International
Maritime
Dangerous Goods
Code (IMDG
Code), ensuring
proper packaging,
marking, and
stowage
5
3 METHODOLOGY
The overall workflow of the proposed methodology is
illustrated in Figure 1.
Figure 1. Methodology and Applications
3.1 BN model
According to Pearl, J. (2014) and Korb and Nicholson
(2010), a BN is a directed acyclic graph (DAG) model
based on probability theory, used to represent
dependencies among variables. Nodes represent
random variables, and edges represent conditional
dependencies. The network makes inferences using
conditional and joint probabilities. Reasoning in a BN
relies on Bayes' theorem, which updates the probability
of an event A occurring, given that event B has
occurred. This is done by calculating the posterior
probability P(A|B).
( )
( ) ( )
( )
|
|
P B A P A
P A B
PB
=
(1)
3.2 EM Algorithm
According to Dempster et al. (1977) and Borman (2004),
the Expectation-Maximization (EM) algorithm is an
iterative method for estimating parameters in
probabilistic models with latent variables or missing
data. It alternates between the Expectation step (E-step)
and the Maximization step (M-step) to optimize
parameters until convergence.
In the E-step, the algorithm calculates the expected
values of latent variables based on current parameter
estimates, which are then used for the next update.
( )
(
)
( )
( )
, log ,
t
t
Z X P X Z
QE




=
∣
(2)
In the “Maximization step (M-step)”, the
parameters are updated by maximizing the expected
values obtained in the E-step. The update formula is:
( ) ( )
(
)
1
argmax
tt
Q

+
=
(3)
This process is iterated repeatedly until the
parameters converge. The final parameter estimates
obtained are the optimal parameters for the BN model.
3.3 Sensitivity Analysis
According to Saltelli et al. (2008) and Oakley and
O'Hagan (2004), sensitivity analysis is used to evaluate
the impact of numerical parameters in a BN (such as
prior probabilities and conditional probabilities) on the
posterior probability. The core idea is to quantify the
sensitivity of the posterior probability to changes in
each parameter by calculating the derivative of the
posterior probability with respect to each parameter.
This measures how changes in the parameters affect
the posterior probability.
Sensitivity is assessed by the absolute value of a
parameter's derivative. A larger derivative indicates
high sensitivity, while a smaller one suggests low
sensitivity. Sensitivity analysis helps identify key
parameters impacting the posterior probability,
guiding model optimization and parameter
adjustments.
3.4 Visualization of Influence Strength Between Nodes
According to Chen et al. (2014) and Korb and
Nicholson (2010), based on Euclidean distance, in a BN,
the thickness of the arcs between nodes represents the
strength of influence, with thicker arcs indicating
stronger influence. Euclidean distance measures the
difference between two probability distributions, and
its formula is:
( )
( )
2
1
,
n
ii
i
d P Q P Q
=
=−
(4)
Here, Pi and Qi are the probabilities of distributions
P and Q at the i-th state, and n is the total number of
states.
4 IDENTIFICATION AND CLASSIFICATION OF
RIFS
4.1 Data Sources
The data for this study comes from 3,792 detailed
accident reports (from 1973 to 2024) available on the
International Maritime Organization (IMO) website.
These reports include information such as the names of
the involved vessels, IMO numbers, accident severity,
occurrence time, location, and accident summaries.
First, by using keyword searches, 158 reports related to
hazardous cargo vessels were initially selected from
the 3,792 reports. Then, by removing irrelevant
content, 106 accident reports specifically related to
hazardous cargo vessels were finalized.
4.2 RIFs
Through the analysis of hazardous chemicals
literature, common RIFs were first identified. Next,
through the analysis of accident reports, additional
RIFs not mentioned in the literature were incorporated.
Finally, the RIFs related to the navigational safety of
hazardous cargo vessels were summarized.
This section summarizes the Risk Influencing
Factors (RIFs) from five aspects: human factors, ship
factors, cargo factors, environmental factors, and
management factors. They are presented in a
hierarchical structure in Figure 2.
6
Figure 2. RIFs classification framework
These factors constitute the nodes of the BN. Below
is a detailed explanation of these factors.
Management Factors include Inadequate Safety
Management (A1), which refers to the absence of a
systematic safety management system on the ship,
resulting in the failure to effectively identify, assess,
and control potential safety risks; Non-compliance
with Navigational Regulations (A2), which involves
failure to operate the ship in accordance with
international or regional navigational rules and
standards; Improper Equipment Maintenance (A3),
characterized by the lack of regular maintenance and
inspection of ship equipment, leading to equipment
failure or performance degradation; Inadequate Safety
Inspections (A4), which denotes the failure to conduct
comprehensive and detailed safety inspections before
or during ship operations; Delayed Emergency
Response (A5), referring to the failure to react
promptly during emergencies, delaying necessary
measures and worsening accident outcomes; and
Inadequate Hazardous Chemicals Management (A6),
which involves poor management of hazardous
chemicals, including improper storage, transportation,
and record-keeping, increasing the risk of leaks or
pollution.
Cargo Factors include Flammable and Explosive
Cargo (D1), which refers to the transportation of cargo
with flammable and explosive characteristics, such as
petroleum, posing risks of fires or explosions if leaked
or exposed to a fire source; and Polluting and Toxic
Cargo (D2), which involves the transportation of cargo
with polluting or toxic properties, potentially causing
serious hazards to the environment and human health
if leaked.
Ship Factors include Improper Cargo Stowage (E1),
which refers to the incorrect loading and securing of
cargo on the ship, leading to instability of the ship's
center of gravity, cargo shifting, or structural damage;
Unseaworthy Condition (E2), which denotes the
overall condition of the ship failing to meet seaworthy
standards, including structural damage, equipment
failure, or poor maintenance, rendering safe navigation
impossible; and Ship Type (E3), which relates to the
design and structural characteristics of the ship,
influencing its safety in specific environments or when
transporting particular types of cargo.
Environmental Factors include Visibility (F1),
which refers to low visibility conditions such as fog,
nighttime, or rainy weather, affecting the ship's
navigation and lookout, thereby increasing the risk of
collision or grounding; Rain (F2), which involves
rainfall that may reduce visibility and worsen
navigational conditions, while also increasing the
difficulty and risk of deck operations; Wind Force (F3),
which denotes strong winds that impact the ship's
maneuverability and stability, increasing navigation
difficulty and potentially leading to loss of control; and
Ocean Waves (F4), which refers to high waves and
adverse sea conditions that exert pressure on the ship's
structure and operations, increasing the risk of
capsizing, grounding, or other accidents.
Human factors include several critical aspects.
Operational errors (G1) involve mistakes made by the
crew during equipment operation or navigation, such
as improper use of instruments, which may lead to
accidents. Violation of safety regulations (G2) refers to
the failure to adhere to safety procedures, such as
neglecting protective gear or skipping inspections,
heightening accident risk. Insufficient training (G3)
reflects a lack of necessary skills, leading to improper
operations or delayed emergency responses. Decision-
making errors (G4) arise when incorrect choices are
made during navigation or emergency handling, such
as selecting wrong routes or ineffective measures.
Failure to maintain a proper lookout (G5) results in the
crew's inability to spot obstacles or other vessels in
time, increasing collision risk. Communication errors
(G6) arise from poor interaction among crew members
or with shore control, leading to misunderstandings
and mistakes. Fatigue (G7) from long working hours or
inadequate rest impairs attention and response time.
Inappropriate emergency responses (G8) occur when
crew members fail to follow established procedures in
critical situations, worsening outcomes. Inadequate
protective measures (G9) reflect a lack of safety
equipment or preparedness, hindering effective
responses to emergencies. Finally, insufficient
knowledge of hazardous chemicals (G10) refers to the
crew's lack of expertise in handling, storing, or
7
responding to hazardous substances, raising the
potential for accidents or contamination.
The ship type (C) has been mentioned earlier, and
the accident types (H) include the following five
categories: Collision (B1), which refers to physical
contact between the ship and other objects, leading to
structural damage or functional failure; Grounding
(B2), which occurs when the ship touches the bottom in
shallow waters, resulting in the inability to navigate
normally; Sinking (B3), which involves the ship losing
buoyancy due to various reasons and eventually
sinking to the bottom; Personnel Poisoning (B4), which
denotes crew members experiencing health issues due
to exposure to toxic substances; and Fire/Explosion
(B5), which refers to fire or explosion incidents on the
ship, potentially leading to severe damage and
casualties.
Further, the risk factor nodes are categorized into
two states: "Yes" and "No." "Yes" indicates that the risk
factor node occurred in a particular accident, while
"No" indicates that it did not occur.
5 RISK ANALYSIS OF HAZARDOUS CARGO
VESSELS NAVIGATION BASED ON BN
5.1 Construction of BN for Navigation Risk of Hazardous
Cargo Vessels
Based on the identified RIFs and their causal
relationships in the reports, a BN is constructed to
reflect the interactions among various RIFs during the
navigation of hazardous chemicals vessels. The
network nodes include factors such as personnel,
ships, cargo, environment, and management, along
with additional nodes for accident types and accident
consequences. Figure 3 presents the BN diagram for
the navigation risks of hazardous cargo vessels.
Figure 3. BN of navigation risk for hazardous chemicals
vessels
5.2 Determination of Conditional Probability Tables
According to equations (2) and (3), appropriate initial
parameter values (θ
(0)
) are selected. In the E-step, based
on the current parameters (θ
(t)
), the posterior
probability distribution of the latent variables (Z) is
computed, which represents the conditional
probability of the latent variables given the observed
data (X) and the parameters (θ
(t)
). Then, in the M-step,
the posterior probability obtained from the E-step
(θ
(t)
(Z)) , is used to maximize the likelihood function,
updating the model parameters (θ) and adjusting the
CPT. Through repeated iterations of the E-step and M-
step, the process continues until the parameter change
is below the preset threshold (0.001) or the maximum
number of iterations (100) is reached, ensuring that the
algorithm converges to stable parameter estimates.
Finally, the CPT is obtained through the EM algorithm
and displayed in the nodes, as shown in Figure 4.
Figure 4. BN after parameter learning
6 ANALYSIS OF RESULTS
6.1 Identification of Key Risk Nodes
Sensitivity analysis is conducted to identify the key
RIFs and the most influential input parameters. Based
on the two most prevalent accident types, B4 (24%) and
B5 (21%), which correspond to "personnel poisoning"
and "fire/explosion," respectively, sensitivity analysis
is performed.
Figure 5. Accident type sensitivity analysis chart
Figure 5 shows the BN after sensitivity analysis
with "accident type" as the target node. The darker the
color, the greater the influence of each factor on the
target node. The results indicate that "Inadequate
Safety Inspections" and "flammable and explosive
cargo" are the highest-impact factors. The loading
quantity of flammable and explosive cargo also
significantly affects accident occurrence, increasing the
probability of fires and explosions.
"Inadequate Hazardous Chemicals Management,"
"polluting and toxic cargo," and "ship type" are
secondary high-impact factors. Inadequate
management of hazardous materials often leads to
leaks of toxic substances. "Pollution and toxic cargo"
pose higher risks during transit, especially when
leakage or operational mistakes occur, leading to
poisoning and fatalities. Certain ship types, such as oil
tankers, chemical tankers, liquefied gas carriers, and
chemical bulk carriers, are more prone to specific
accidents. For example, chemical tankers are more
likely to cause poisoning, fires, and explosions due to
the hazardous nature of the cargo, while oil tankers are
typically involved in oil spills after collisions, causing
environmental pollution or casualties.
8
Figure 6. Sensitivity tornado chart when H is B4
Figure 6 shows the results of sensitivity analysis for
the "accident type" of "personnel poisoning" (H = B4).
As seen in the figure, the most significant parameter
influencing "personnel poisoning" is "D2 Pollution and
Toxic Cargo." The current value of this parameter is
0.59434, with a range between 0.416038 and 0.772642,
indicating that the current management of pollution
and toxic cargo is at a moderate risk level. Within this
range, as the proportion of pollution and toxic cargo
increases, the probability of "personnel poisoning"
rises significantly, with a derivative value of 0.0810969,
suggesting that this parameter has a very strong impact
on the target state.
The "ship type" parameter for "liquid chemical
tanker" (E3 = C2) also has a considerable effect. Its
current value is 0.54717, with a range between 0.383019
and 0.711321. This range indicates that an increase in
the proportion of liquid chemical tankers raises the risk
of personnel poisoning, particularly as the ship type
shifts toward liquid chemical tankers. The current
value lies in the middle of this range, indicating a
moderate level of influence on personnel poisoning,
with a derivative value of 0.0463917, showing a
moderate degree of impact.
The combined effect of parameters "G3 Insufficient
training" and "A5 Delayed Emergency Response" has a
derivative value of 0.0305763, indicating that the
interaction between these two factors significantly
increases the probability of "personnel poisoning."
Specifically, insufficient personnel training and
delayed emergency response reduce the efficiency of
critical emergency measures during an emergency,
thereby increasing the risk of personnel poisoning.
From the sensitivity analysis, it can be concluded
that "pollution and toxic cargo" and "liquid chemical
tanker" type are the primary factors influencing
"personnel poisoning" accidents. Additionally,
"insufficient personnel training" and "delayed
emergency response" significantly increase the
poisoning risk.
Figure 7. Sensitivity tornado chart when H is B5
Figure 7 presents the tornado diagram for the
sensitivity analysis when the "accident type" is
"fire/explosion." From the diagram, it is clear that the
most sensitive parameter node is "flammable and
explosive cargo," with a current value of 0.54717 and a
range between 0.383019 and 0.711321. This means that
when the quantity of flammable and explosive cargo
increases or its proportion on the ship grows, the target
state (the probability of fire and explosion) will
fluctuate between 0.198162 and 0.21243, significantly
increasing the probability of fire and explosion
incidents on the vessel.
Next, "Inadequate safety inspections" is also a
highly sensitive parameter node, with a current value
of 0.603774 and a range between 0.422642 and 0.784906.
The influence of this parameter on the target node
ranges from 0.200876 to 0.209715, mainly showing a
positive effect, with a derivative value of 0.024398. This
suggests that as the degree of "Inadequate safety
inspections" increases, the probability of fire and
explosion will correspondingly rise.
In the third analysis of the tornado diagram, when
"A6 = yes" and "A4 = yes," that is, in the case of
"Inadequate Hazardous Chemicals Management" and
"Inadequate safety inspections," the current value is
0.796875, with a range from 0.557812 to 1. This
combination has an influence range on the target node
from 0.201046 to 0.208907, with a derivative value of
0.0177763. This indicates that when "Inadequate
hazardous materials management" and "Inadequate
safety inspections" act together, the risk of fire and
explosion on the vessel is exacerbated.
The above analysis demonstrates that "flammable
and explosive cargo" and "Inadequate safety
inspections" are key factors influencing fire and
explosion accidents on ships. Furthermore, Inadequate
hazardous materials management, in the context of
insufficient safety inspections, further exacerbates the
risk of such accidents.
6.2 Analysis of Causation Chains of Key Risk Nodes
Figure 8 presents the inter-node impact strength
diagram based on maximum values. The results show
that "Inadequate safety inspections" and "Inadequate
personnel training" are the core nodes of risk
propagation, followed by "Inadequate safety
management," "communication errors," "operational
errors," and "Unseaworthy Condition."
9
Figure 8. Graph based on the maximum node influence
strength
Specifically, "Inadequate safety inspections"
significantly amplifies system risk through multiple
pathways, particularly its notable impact on
"Inadequate hazardous materials management"
(average value: 0.785247), directly causing
management gaps and increasing potential risks.
"Inadequate personnel training" significantly affects
"insufficient knowledge of hazardous chemicals"
(average value: 0.484756), "operational errors" (average
value: 0.212221), and "Inadequate protective measures"
(average value: 0.242708), reflecting the constraint that
the lack of training places on personnel's operational
capabilities and awareness of safety measures.
"Inadequate safety management" has a destructive
impact on system stability by promoting "Failure to
maintain a proper lookout" (average value: 0.358623)
and non-compliance with navigation rules.
"Communication errors" significantly influence
"operational errors" (average value: 0.210832) and
"Decision-making errors" (average value: 0.0712318),
indicating that the efficiency of information transfer
and collaboration is critical to risk control. "Operational
errors," as direct triggers of accidents, are involved in
multiple risk pathways, further increasing the
likelihood of accidents.
"Unseaworthy Condition" integrates multiple
factors such as equipment maintenance, navigational
adaptability, and external environment, significantly
contributing to increased accident risks. Additionally,
"accident type" encompasses various situations,
including collisions, groundings, sinking, personnel
poisoning, fire, and explosion, and serves as a
concentrated output point for multiple risk nodes,
reflecting the interactions of various RIFs. Among
these, each accident results in property damage, with
the highest impact value being 0.4375; personnel
casualties are mainly caused by poisoning incidents
due to fire, explosion, and Inadequate protective
measures; environmental pollution typically arises
from ship collisions or groundings that damage the
hull, leading to cargo leakage or loss.
In summary, the key risk causality chains include:
1. Insufficient training → Insufficient knowledge of
hazardous chemicals → Operational errors →
Accident type
2. Inadequate safety inspections → Inadequate
hazardous materials management →
Environmental pollution
3. Communication errors → Operational errors →
Accident type
4. Unseaworthy Condition → Operational errors →
Accident type
These key risk nodes reveal that insufficient
training, poor communication mechanisms, and weak
ship maintenance management are the primary causes
of high-risk events.
7 CONCLUSION
This study constructs a BN model for the navigation
risks of hazardous cargo vessels to identify key RIFs
and determine the primary risk paths. The research
found that "inadequate personnel training,"
"Inadequate safety inspections," "flammable and
explosive cargo," "inadequate hazardous materials
management," and "polluted and toxic cargo"
significantly impact maritime accidents. Based on these
key RIFs and their evolution, it aims to provide
theoretical support for the navigation safety of
hazardous cargo vessels.
Future research can further explore additional
potential RIFs based on further data collection and
conduct long-term tracking and evaluation of the
effectiveness of the risk control measures, aiming to
optimize hazardous chemicals vessels navigation
safety management strategies.
ACKNOWLEDGEMENT
This work is supported by the National Key R&D Program of
China (2024YFB4303600), and the National Natural Science
Foundation of China (52572386, 52425210).
REFERENCES
Borman, S. (2004). The expectation maximization algorithm-
a short tutorial. Submitted for publication, 41.
Cao, Y., Wang, X., Wang, Y., et al. (2023). Analysis of factors
affecting the severity of marine accidents using a data-
driven Bayesian network. Ocean Engineering, 269,
113563.
Chen, P., Zhang, Z., Huang, Y., et al. (2022). Risk assessment
of marine accidents with Fuzzy Bayesian Networks and
causal analysis. Ocean & Coastal Management, 228,
106323.
Chen, Y., & Tian, J. (2014). Finding the k-best equivalence
classes of Bayesian network structures for model
averaging. Proceedings of the AAAI Conference on
Artificial Intelligence, 28(1).
Dempster, A. P., Laird, N. M., & Rubin, D. B. (1977).
Maximum likelihood from incomplete data via the EM
algorithm. Journal of the Royal Statistical Society: Series
B (Methodological), 39(1), 1-22.
Elidolu, G., Teixeira, Â. P., & Arslanoğlu, Y. (2024). A risk
assessment of inhibited cargo operations in maritime
transportation: A case of handling styrene monomer.
Ocean Engineering, 312, 119049.
Fan, S., Zhang, J., Blanco-Davis, E., et al. (2020). Maritime
accident prevention strategy formulation from a human
factor perspective using Bayesian Networks and TOPSIS.
Ocean Engineering, 210, 107544.
Feng, Y., Liu, Z., Jiang, Z., et al. (2023). Analysis of factors
affecting ship collisions based on association rule mining
and complex network theory. Dalian Haishi Daxue
Xuebao/Journal of Dalian Maritime University, 49(3), 31-
44.
Guo, J., Luo, C., & Ma, K. (2023). Risk coupling analysis of
road transportation accidents of hazardous materials in
10
complicated maritime environment. Reliability
Engineering & System Safety, 229, 108891.
Hänninen, M., Banda, O. A. V., & Kujala, P. (2014). Bayesian
network model of maritime safety management. Expert
Systems with Applications, 41(17), 7837-7846.
International Maritime Organization. (2025). International
Maritime Dangerous Goods Code (IMDG Code). 42-24
Edition. London: International Maritime Organization
Khan, R. U., Yin, J., Ahani, E., et al. (2024). Seaport
infrastructure risk assessment for hazardous cargo
operations using Bayesian networks. Marine Pollution
Bulletin, 208, 116966.
Khan, R. U., Yin, J., Mustafa, F. S., et al. (2021). Risk
assessment for berthing of hazardous cargo vessels using
Bayesian networks. Ocean & Coastal Management, 210,
105673.
Korb, K. B., & Nicholson, A. E. (2010). Bayesian artificial
intelligence. CRC Press.
Ma, L., Ma, X., Liu, Y., et al. (2023). Risk assessment of
coupling links in hazardous chemicals maritime
transportation system. Journal of Loss Prevention in the
Process Industries, 82, 105011.
Oakley, J. E., & O'Hagan, A. (2004). Probabilistic sensitivity
analysis of complex models: a Bayesian approach. Journal
of the Royal Statistical Society Series B: Statistical
Methodology, 66(3), 751-769.
Pearl, J. (2014). Probabilistic reasoning in intelligent systems:
networks of plausible inference. Elsevier.
Qiao, W., Liu, Y., Ma, X., et al. (2020). Human factors analysis
for maritime accidents based on a dynamic fuzzy
Bayesian network. Risk Analysis, 40(5), 957-980.
Saltelli, A. (2008). Global Sensitivity Analysis: the Primer.
John Wiley & Sons.
Sezer, S. I., & Akyuz, E. (2024). A conceptual risk modelling
for cargo tank fire/explosion in chemical tanker by using
Evidential Reasoning-SLIM and Bayesian belief network
approach. Reliability Engineering & System Safety, 252,
110455.
Sezer, S. I., Elidolu, G., Akyuz, E., et al. (2023). A quantified
risk analysis for oil spill during crude oil loading
operation on tanker ship under improved Z-number
based Bayesian Network approach. Marine Pollution
Bulletin, 197, 115796.
Shi, Weihua. (2017). Study on Risk of Marine Transportation
of Dangerous Chemicals Based on Safety Engineering
Reliability. International Journal of Physical and Social
Sciences, 7(4), 52-63.
Wu, J., Bai, Y., Zhao, H., et al. (2021). A quantitative LNG risk
assessment model based on integrated Bayesian-
Catastrophe-EPE method. Safety Science, 137, 105184.
Yang, Z., Wan, C., Yang, Z., et al. (2021). Using Bayesian
network-based TOPSIS to aid dynamic port state control
detention risk control decision. Reliability Engineering &
System Safety, 213, 107784.
Zhang, D., Yan, X. P., Yang, Z. L., et al. (2013). Incorporation
of formal safety assessment and Bayesian network in
navigational risk estimation of the Yangtze River.
Reliability Engineering & System Safety, 118, 93-105.
Zhang, X., Zhang, Q., Yang, J., et al. (2019). Safety risk
analysis of unmanned ships in inland rivers based on a
fuzzy Bayesian network. Journal of Advanced
Transportation, 2019(1), 4057195.