Risk Analysis of Pile Pitching and Pulling on Offshore Wind Power Jack-Up Platforms Based on a Fault Tree and Fuzzy Bayesian Network

Abstract

Safety accidents during pile pitching and pulling operations on offshore wind power jack-up platforms occur frequently, yet research into their underlying causes is insufficient. This study delved into the causes of accidents related to pile pitching and pulling and put forward corresponding risk prevention and control measures by integrating the Fault Tree Analysis (FTA) and Fuzzy Bayesian Network (FBN) in consideration of the high-risk characteristics of these operations. Firstly, this study expounded the causal relationship of risk factors in the pile pitching and pulling operations on offshore wind power jack-up platforms via FTA. Secondly, the events in the FTA model were mapped to the FBN nodes. The prior probabilities of each node were determined through expert evaluation, and a Fuzzy Bayesian Network model was constructed. Finally, risk diagnosis and prediction were carried out through probability updating and a sensitivity analysis. The results indicate that environmental risks, including water depth, strong winds, heavy waves, and unknown subsea geology, exert the most significant influence. Equipment malfunctions and management problems are the key causes of accidents. A sensitivity analysis reveals that failures in the pile driving system and underwater monitoring system are highly sensitive triggers for the top-level event. Improvement measures are proposed to mitigate risks and enhance project safety.

1. Introduction

As the demand for energy resources and the emission of greenhouse gases increase, multiple countries have taken the enhanced application in clean energy to be the primary goal of sustainable development strategies [1]. China embraces abundant offshore wind energy resources [2], with the total installed capacity of offshore wind power reaching 41.27 million kilowatts by the end of 2024, ranking first worldwide for four consecutive years. By 2025, China is expected to add an installed capacity of 30–40 million kilowatts of offshore wind power [3]. However, the construction of offshore wind power shows an upward trend in the accident rate despite its unprecedented development [4]. The 2023 safety report, published by the Global Offshore Wind Health and Safety Organization G+, along with pertinent industry data, indicates that there were 1679 safety incidents within the global offshore wind industry last year, marking a 94% increase from the 867 incidents reported in 2022. Among them, the top three operations with the highest number of accidents are lifting operations (207), vessel operations including jack-up and barge operations (169), and daily maintenance operations (109). The operation of pile pitching and pulling on the offshore wind power jack-up platform is of high risk and prone to cause safety accidents due to risk factors including complex and variable seabed geological conditions, dynamic impacts of marine environment, equipment failure, and insufficient safety management [5]. On 25 July 2021, during a pile installation operation at an offshore wind farm located at Huizhou Port, China, the offshore wind power jack-up platform “Shengping 001” suffered a hull puncture caused by a pile leg, leading to structural tilting and subsequent sinking. The incident resulted in one fatality, and three missing personnel. Nevertheless, the analysis of the causes behind operations involving the pile pitching and pulling on offshore jack-up platforms is currently quite limited. This scarcity is primarily attributed to the limitations of existing research and the challenges associated with accident data. Therefore, managing safety risks associated with pile pitching and pulling on the jack-up platform is of significant importance.

Offshore jack-up platforms are versatile, capable of operating in various seabed geological conditions, large water depths, and differing marine gas-phase environments. They also play a crucial role in the construction of offshore wind farms, providing a stable platform for the installation and maintenance of wind turbines, thereby facilitating a range of construction activities for personnel [6]. Pitching and pulling are the core processes involved in deploying and evacuating offshore platforms. They encompass several complex steps, including geological exploration, environmental monitoring, precise platform positioning, preloading of pile legs, platform fixation, and lifting. These processes present significant technical challenges and numerous potential risks. Common accidents during pile pitching and pulling include pile leg piercing, pile sliding, structural failures, and platform overturning. These incidents can result in a large number of casualties and substantial equipment damage upon occurrence. Currently, researchers have developed a series of models focusing on technical enhancements to tackle issues like penetration failure in layered soil under large 3D jack-up platforms, pile shoe penetration accidents in sand-over-clay layers, simulation of pile shoe installation and the subsequent consolidation–settlement in clay seabeds, and the evaluation of jack-up platforms’ overall response and inter-pile shoe load distribution in sand-over-clay formations. Phuor et al. [7] proposed a new algorithm that combines the element-wise preconditioned conjugate gradient method with the interface model of oblique boundary conditions, the load control algorithm based on generalized stiffness parameters, and the soil viscoplastic model to effectively handle the penetration failure of layered soil on a large-scale, three-dimensional jack-up platform. Yin et al. [8] conducted large-scale field experiments and systematic studies on the penetration process of pile shoes on the jack-up platform under geological conditions of sand-covered clay layers and proposed a new plan to prevent penetration accidents, thereby providing technical guidance for practical engineering operations. Yi et al. [9] proposed a finite element method that integrates the Euler–Lagrange two-stage technique and successfully simulated the installation and subsequent consolidation and settlement process of pile shoes on the jack-up platform in a clay seabed. This method now provides a unified prediction model for evaluating the consolidation and settlement of pile shoes. Wang et al. [10] proposed a macroscopic unit model of pile shoes suitable for geological conditions of sand-covered clay layers and demonstrated its significant advantages in evaluating the overall response of jack-up platforms and load distribution of pile shoes through a quasi-static overturning analysis. Despite the strong technical support of these studies for the safe construction and accident prevention of pile pitching and pulling on the offshore wind power jack-up platform, there is still a lack of relevant cause analysis, which mainly derives from the limitations of existing research and the difficulty in data collection. Owing to the paucity of data on the fundamental risk factors of offshore wind power engineering construction operations, quantifying the risk analysis of such operations has emerged as a formidable task. With respect to the safety concerns of the pile pitching and pulling construction operations of the offshore wind self-elevating platform, the risk indicators throughout the construction process of the offshore wind self-elevating platform’s pile pitching and pulling were systematically organized. Based on this, approaches to address the information uncertainty arising from the fuzzy and interdependent relationships among multiple factors were put forward. A dynamic risk evolution model for the pile pitching and pulling construction operations of the offshore wind self-elevating platform under uncertain conditions was established, which constitutes the core algorithm for the dynamic risk assessment of these operations.

An in-depth analysis of the causes of pile pitching and pulling on offshore wind power jack-up platforms is crucial for optimizing operational processes and standards, as well as enhancing operational safety. The installation and operation of pile pitching and pulling on the offshore jack-up platform are encountering many potential causes. Platforms are typically deployed in underwater environments with complex geological conditions, especially in complex foundation conditions such as soft clay-covered sand layers. Insufficient exploration of geological conditions or inadequate preloading operations are prone to “piercing” the platform during installation; that is, the platform penetrates the overlying sand layer and enters the underlying soft clay layer, leading to a sharp decrease in the foundation bearing capacity and posing a serious threat to the overall stability of the platform [11]. The environmental load is also an important risk factor for the platform. Extreme weather conditions, such as storms, huge waves, and strong ocean currents, can exert enormous horizontal forces, bending moments, and vertical loads on the platform, leading to structural overload or foundation damage [12]. In addition, management factors cannot be ignored. Operator errors or negligence, such as incorrect operating procedures and the failure to promptly detect and repair structural damage, may also lead to safety accidents. Scientific and systemic cause analyses act as key steps in accident prevention [13]. At present, there are various methods widely used for cause analysis on accidents, such as the Fault Tree Analysis (FTA), Fault Mode and Effect Analysis (FMEA), Bayesian Network (BN), Systems-Theoretic Accident Model and Processes (STAMP), etc. However, the single use of these methods has its limitations. In this case, researchers have proposed multiple integrated methods. FTA and FBN can be combined to conduct the fuzzy processing of subjective data and enhance structure–probability synergy. They have become a powerful tool in cause analysis on accidents and have been widely used in multiple fields [14], with the most frequent application in the field of maritime transportation. By integrating FTA and FBN, modifying the Cognitive Reliability and Error Analysis Method (CREAM), and combining accident databases, expert judgments, and the fuzzy set theory, Ung et al. [15] revealed that inadequate communication, crew fatigue, unauthorized operations, and multi-tasking on bridges (CPC4) are the main factors in collisions. On this basis, a human error assessment framework capable of handling fuzzy data was established to provide a scientific basis and decision support for marine safety. Li et al. [14] combined the FTA and BN to systematically evaluate the impact of human factors on serious maritime accidents. Muhammet Aydin et al. [16] conducted a comprehensive assessment of fire and explosion risks in the cargo hold and deck of bulk carriers using FBN; identified key risk factors such as thermal work, cable insulation damage, and unclosed cargo hold lights; and quantified the contribution of these factors to the occurrence of accidents. The research results are conducive to improving the safety management level of bulk carriers, reducing fire and explosion accidents, and ensuring the safety of personnel and the environment. Tuncel et al. [17] conducted a detailed analysis on the risk of fire and explosion accidents in bulk carriers using a fuzzy FTA and revealed the main risk factors and their most effective accident combinations, thereby providing a scientific basis for reducing risks and improving operational safety. FTA and FBN are widely used in cause analyses on accidents due to their strong data acquisition and processing capabilities as well as high inference efficiency, thereby providing strong support for accident prevention and safety management [18]. The construction process of the self-elevating platform for offshore wind power encompasses a multitude of procedures. The marine environment is characterized by complexity and variability, and the uncertain factors influencing operational safety are substantial. The identification and management of construction risks encounter significant challenges. This paper employs the Fuzzy Bayesian method to systematically organize and establish the full-process risks of the self-elevating platform for offshore wind power with pile pitching and pulling. It constructs a multi-dimensional dynamic scenario identification method for the construction risks of the self-elevating platform for offshore wind power with pile pitching and pulling, determines the risk priority order of each identification unit, and achieves a more comprehensive identification and a more thorough evaluation of potential risk sources.

In this study, a causal chain for accidents was constructed by combining the Fault Tree Analysis (FTA) and Fuzzy Bayesian Network (FBN) to effectively enhance the risk-resistance capability of offshore wind power construction projects and reduce or eliminate accidents caused by pile pitching and pulling on the offshore jack-up platform. The FTA-FBN can effectively address the issues of subjectivity and data loss in personnel evaluation, thereby overcoming the limitations of traditional methods that rely solely on accurate data to uncover the causes of accidents. Meanwhile, the probabilistic reasoning capability of Bayesian Networks (BNs) enables the quantitative analysis of dynamic relationships within complex systems, using the structured framework of accident trees. This not only preserves the logical clarity of causal paths but also compensates for the shortcomings of single methods in nonlinear modeling, quantitative decision-making, and practicality. Especially in scenarios with limited data or where expert knowledge is dominant, it demonstrates significant advantages by identifying key risk paths through accident tree models and quantifying the correlations between uncertain risk factors using Bayesian Networks (FBNs). The safety and reliability of pile pitching and pulling on the offshore wind power jack-up platform can be improved by thoroughly analyzing the risk factors during these processes, accurately evaluating the risks, identifying critical risk factors, and formulating effective risk control measures. This approach is highly significant for ensuring personnel safety, minimizing economic losses, and facilitating the sustainable development of the offshore wind power industry. In the risk management process of offshore wind power engineering construction operations, it is imperative to initially control and diminish the occurrence probability of the identified risk factors. This approach can enhance operational reliability and mitigate risks throughout the entire construction operation process. The dynamic Bayesian assessment method is employed to dynamically evaluate the operation procedures of the offshore wind self-elevating platform, discern the operational safety situation, offer a foundation for the judgment of the operational safety status, and effectively identify the key safety risk points, thus providing guidance for efficient and secure management.

2. Methods

Figure 1 shows the specific steps for combining FTA and FBN, which can efficiently evaluate complex system risks by integrating fault logic analysis and probability uncertainty, especially suitable for scenarios where data is missing or dynamic risk analysis relies on expert experience [19]. Firstly, risk factors of pile pitching and pulling on the offshore jack-up platform were identified based on FTA, and the top events were decomposed layer by layer through logic gates (such as AND gates, OR gates, etc.) to the basic events that cannot be further divided. Secondly, the events (top event, middle event, and basic event) of the developed FTA model were mapped to the nodes of a Bayesian Network (BN), and the prior probabilities of each node in the BN were determined using expert evaluation and fuzzy sets (such as “very low”, “low”, “medium”, “high”, and “very high”). Finally, the posterior probability was obtained using a Bayesian Network inference algorithm, and sensitivity analysis was conducted to further identify the key risk factors of pile pitching and pulling on the offshore jack-up platform and to optimize risk control measures.

2.1. Fault Tree Analysis (FTA)

Fault Tree Analysis (FTA) [20] employs logical reasoning to visually depict the causal relationships of accidents in a graphical form. A Fault Tree is a directed acyclic network consisting of two distinct types of nodes: events and gates. The Fault Tree commences from the top-level accident (top event) and links various causal events via logic gates (e.g., AND gates, OR gates, etc.), tracing back to the basic events of the accident [21]. Basic events are initial faults or errors that require further decomposition. They are situated at the bottom of the Fault Tree and are typically represented by a circular pattern. In contrast, top events or intermediate events are fault events that occur subsequent to the logical combination of basic events, encompassing system states, and are generally represented by a rectangular pattern [22]. All these events are interconnected using traditional logic gates, combined with the logical connections of AND and OR gates, visually presenting the propagation path of accidents from underlying faults to top-level consequences. This, in turn, offers a clear framework for systemic safety analysis [23]. Regarding accidents, their immediate causes can be analyzed, and underlying causes can be thoroughly uncovered. Consequently, the critical path and weak links of the accident can be identified to develop effective preventive measures, thereby improving the safety of the system.

2.2. Bayesian Network (BN)

Bayesian Networks possess a robust capacity to address uncertainty problems and execute learning and reasoning under conditions of limitations, incompleteness, and uncertainties [24,25]. Commonly employed to represent a set of variables and their conditional dependencies, Bayesian Networks are composed of a directed acyclic graph (DAG) and a conditional probability distribution (CPD), which can visually present the causal and probabilistic relationships among variables [26]. In Bayesian Networks, nodes signify random variables, edges denote direct dependencies between variables, and the direction of edges points from the parent node to the child node. Each node is associated with a conditional probability distribution to depict the probability under the condition of known parent nodes [27].

The quantitative calculation component of Bayesian Networks is primarily accountable for processing the probabilities of various variables within the network, encompassing three key elements: prior probability, conditional probability, and posterior probability. The prior probability and conditional probability in the network can be acquired through direct analysis of historical data or expert assessment. Essentially, Bayesian theory defines the probability of an event as a priori knowledge contingent upon conditions potentially relevant to the event [25].

2.3. Fuzzy Bayesian Network (FBN)

Fuzzy Bayesian method is a technique that combines fuzzy logic and Bayesian inference, mainly used for scenarios that require handling subjective judgments or missing data [28]. The Fuzzy Bayesian method integrates fuzzy set theory into the Bayesian framework, using fuzzy language variables such as “very low, low, medium, high, and very high” to describe the likelihood of an event and quantifies it through triangular or trapezoidal fuzzy numbers [29], making it closer to the uncertainty problems in reality. The core lies in using membership degrees to express fuzzy knowledge and updating probabilities through the framework of the Bayesian Network method [30,31,32].

As the complexity and technical requirements of the operating sea environment increase, pile pitching and pulling on the offshore jack-up platform are facing huge risks and challenges compared with onshore engineering or shallow water operations. Pile pitching and pulling operations must navigate through various uncertainties, including seabed geological conditions, tidal dynamics, and platform stability. Moreover, traditional analysis methods, which rely on single data sets, struggle to fully uncover the inherent correlations within these operations. This is because the causal chain of accidents often involves multi-source coupling risks, such as geological exploration errors, equipment fatigue damage, and operational errors, among others. Therefore, it is necessary to model the risk evolution mechanism based on the experience of experts in relevant fields, so as to compensate for the limitations of data loss. To this end, this study utilized an FBN model and integrated fuzzy language evaluations from experts on key nodes such as “work environment,” “equipment status,” and “personnel status.” Fuzzy set theory was also utilized to convert these evaluations into fuzzy probability distributions for the root nodes, and Bayesian inference was employed to quantitatively calculate the probabilities of the nodes. Together, this study offers probabilistic inference support for decision-making regarding the risks of pile pitching and pulling on the offshore operation platform. Detailed information on this method is presented below.

2.3.1. Capturing Possibilities from Expert Judgment

In this study, the expert evaluation method was used. The weight of each expert was calculated to ensure the accuracy of the evaluation data, considering that the background and experience of experts can affect their judgments. Specifically, scholars generally believe that factors such as age, professional position, years of experience, and education level are key factors affecting the judgment of experts [33]. These factors are typically used to evaluate expert profiles and decision weights [13,17,25]. Based on their classification and the weight assigned to each factor, as shown in

Table 1. Standards and scores of expert weight.

Standards	Classification	Scores
Professional positions	Professorate Senior Engineer/Professor	10
	Senior Engineer/Associate Professor/Manager	8
	Engineer	6
Years of experience	$\ge$20	10
	15–19	8
	10–14	6
	6–9	4
Academic degree	Doctor	10
	Master	8
	Bachelor	6
Age	$\ge$50	8
	40–49	6
	30–39	4

, Formulas (1) and (2) can be used to determine the weight of expert judgments for each root node:

$${\omega }_{\mathrm{j}}={PP}_{\mathrm{j}}+{E\mathrm{d}}_{\mathrm{j}}+{E\mathrm{p}}_{\mathrm{j}}+{Ag}_{\mathrm{j}}$$

(1)

$$W{f}_{\mathrm{j}}={\displaystyle \frac{{\omega }_j}{{\displaystyle \sum _{\mathrm{j}=1}^n{\omega }_j}}}$$

(2)

where ω_j is the weight score of expert j, PP is the professional position, Ed is education, Ep is experience, and Ag is the age of expert j. Wf represents the total weight factor of each expert j.

2.3.2. Fuzzification

Fuzzy set theory is a mathematical theory based on fuzzy mathematics used to describe things that are uncertain, fuzzy, or have incomplete information. This theory holds that an element can belong to multiple sets, and its degree of membership can be represented by a real number between 0 and 1 [19]. This theory holds a core idea of membership functions, which view the relationship between elements and sets as a continuous process rather than discrete boundaries. Fuzzy set theory, consisting of the foundation and extension of fuzzy mathematics, fuzzy logic, fuzzy relationships, fuzzy systems, etc., has been widely applied in multiple fields and has become an important tool for dealing with fuzzy and uncertain problems [28]. In this study, trapezoidal fuzzy numbers were used as the membership function for expert evaluation of language conversion. The membership function x of the trapezoidal fuzzy number X is shown in Formula (3).

$$u\tilde{R}\left(x\right)=\{\begin{array}{ll}0 \hfill & \hfill x<a \\ {\displaystyle \frac{x-a}{b-a}} \hfill & \hfill a\le x\le b \\ 1 \hfill & \hfill b\le x<c \\ {\displaystyle \frac{x-d}{c-d}} \hfill & \hfill c\le x\le d\\ 0 \hfill & \hfill x\ge d \end{array}$$

(3)

In the formula, a, b, c, and d are the four values of the trapezoidal fuzzy number interval. When b = c, trapezoidal fuzzy numbers are transformed into triangular fuzzy numbers.

In the present study, a table of fuzzy numbers associated with five terms from the fuzzy theory was chosen as expert data to assess the probability distribution of node uncertainty.

Table 2. Language terms and fuzzy numbers.

Language Terms	Fuzzy Numbers
VL	(0.0, 0.0, 0.1, 0.2)
L	(0.1, 0.25, 0.25, 0.4)
M	(0.3, 0.5, 0.5, 0.7)
H	(0.6, 0.75, 0.75, 0.9)
VH	(0.8, 0.9, 1.0, 1.0)

presents the fuzzy numbers corresponding to the fuzzy theory terms “very low, low, medium, high, and very high,” which are abbreviated as VL, L, M, H, and VH, respectively [16,32].

2.3.3. Defuzzification

In this study, the Centroid Method (CM) was used to quantitatively score the fuzzy likelihood of uncertain events. By calculating the geometric centroid of fuzzy sets, this method can more comprehensively integrate the distribution characteristics judged by experts to avoid cognitive bias that may be caused by single membership degree truncation, thereby minimizing knowledge loss in the fuzzy reasoning process and providing analytical basis for complex decision-making problems [34,35]. Formulas (3)–(5) can be used to obtain the input of the failure probability of the root node.

The possibility of fuzzy failure can be expressed as

$$Z_i={\displaystyle \sum _{j=1}^n{\omega }_j}{f}_{ij},i=1,2,\dots ,m,j=1,2,\dots ,n$$

(4)

In the formula, Z_i denotes the fuzzy aggregation value of the basic event i, ω_j is the weighting factor of expert j, f_ij is the linguistic value of expert j for the event, M is the total number of basic events, and n is the total number of experts.

For trapezoidal fuzzy sets $\tilde{A}=(a_1,a_2,a_3,a_4)$, the fuzzy possibility X* of basic events can be expressed as

$$X^{*}={\displaystyle \frac{{\displaystyle \underset{a_1}{\overset{a_2}{\int }}\frac{x-a_1}{a_2-a_1}xdx+{\displaystyle \underset{a_2}{\overset{a_3}{\int }}xdx}+{\displaystyle \underset{a_3}{\overset{a_4}{\int }}\frac{a_4-x}{a_4-a_3}xdx}}}{{\displaystyle \underset{a_1}{\overset{a_2}{\int }}\frac{x-a_1}{a_2-a_1}dx+{\displaystyle \underset{a_2}{\overset{a_3}{\int }}dx+{\displaystyle \underset{a_3}{\overset{a_4}{\int }}\frac{a_4-x}{a_4-a_3}dx}}}}}={\displaystyle \frac{1}{3}}\left({\displaystyle \frac{{\left(a_4+a_3\right)}^2-a_3{a}_4-{(a_1+a_2)}^2+a_1{a}_2}{a_4+a_3-a_2-a_1}}\right)$$

(5)

2.3.4. The Calculation of FPr

The core of fuzzy probability (FPr) lies in accurately quantifying the probability distribution of events in a fuzzy context through membership functions and fuzzy set theory [36]. FPr distribution or fuzzy number is transformed into a fuzzy probability of each basic event by merging input uncertainties [37].

$$FP\mathrm{r}=\{\begin{array}{l}{\displaystyle \frac{1}{{10}^k}},X^{*}\ne 0\\ 0,X^{*}=0\end{array},k={({\displaystyle \frac{1-X^{*}}{X^{*}}})}^{{\displaystyle \frac{1}{3}}}\times 2.301$$

(6)

3. Establishment of a Model for Pile Pitching and Pulling on the Offshore Jack-Up Platform—Results

3.1. Establishment of the Fault Tree

Pile pitching and pulling on the offshore wind power jack-up platform involve multiple and complex risk factors. In terms of the environment, strong winds, huge waves, ocean currents, and complex geological conditions can easily cause platform shaking and uneven stress on the pile legs. In terms of equipment, failure in the lifting system, hydraulic devices, and flushing protection devices may lead to abnormal lifting or platform instability. In terms of management, a nonstandard workflow, missing emergency plans, and inadequate safety supervision will increase the risk of accidents. In terms of operation, insufficient professional skills and weak safety awareness among personnel can lead to misoperation. These factors are intertwined and significantly increase the safety risks of pile pitching and pile pulling. In this study, 12 intermediate events and 31 basic events were identified based on the characteristics and actual situation of pile pitching and pulling on the offshore jack-up platform by investigating accident reports and the industry literature in this field and using expert consultation methods. These basic events were logically associated with the 12 intermediate events using the OR(+) logic gate.

Table 3. Safety risk assessment index system for pile pitching and pulling on the offshore jack-up platform.

Symbol	Meaning	Explanation	Symbol	Meaning	Explanation
K	Risks of pile pitching and pulling on the offshore jack-up platform	Potential hazards that may be encountered during the installation, migration, or removing of the platform when its pile legs are inserted or removed from the seabed.	X10	Failure of underwater positioning system	The failure of the underwater positioning system leads to inaccurate positioning information for each pile leg during vessel lifting.
M1	Environmental risks	An uncertain event caused by natural factors that poses a threat to the operating system.	X11	Failure of remote sensing equipment	Failure of control and remote sensing equipment (including water meters, load indicators, inclinometers, and hydraulic pressure gauges, etc.).
M2	Risks of equipment	The operational risk directly caused by the failure of equipment.	X12	Failure of the protective device of the lifting system	Damage or jamming of the pile legs caused by the failure of the safety protection device of the vessel lifting system.
M3	Management risks	Systemic risks caused by management decisions or process defects.	X13	Failure of flushing protection device	Protecting the soil near the pile foundation to avoid erosion failure.
M4	Risks of personnel	Operational risks caused by personnel behavior or status.	X14	Unclear geological exploration before operation	Geological survey of the work area has not been conducted, and precise investigation of the strata and geological structures in the work area has not been carried out.
M5	Water environmental risks	The main causes are excessive wind speed, surging waves, and intense waves.	X15	Unclear pre-pressure inspection before operation	The loading condition of the jack-up platform has not been calculated and checked, and the load on a single pile leg is greater than the maximum jacking load.
M6	Underwater environmental risks	The main underwater environmental risks include unknown seabed geological conditions, obstacles underwater, and changes in water depth.	X16	Unclear equipment inspection before operation	The pile driving system has not been tested before operation, and the condition of the pile legs and the presence of binding objects have not been checked.
M7	Failure of the detection system	Including underwater monitoring systems, underwater positioning systems, and remote sensing equipment.	X17	Inadequate on-site supervision	The management personnel did not supervise the entire process of the operation in real-time on site and promptly handle various problems.
M8	Failure of the protective device	Including safety protection devices for the lifting system and erosion protection devices.	X18	Insufficient management of personnel in the pile fixing room	Unrelated personnel may enter the pile fixing room during platform lifting.
M9	Risk identification before operation	Including geological exploration, pre-pressure inspection, and equipment inspection.	X19	Inadequate personnel qualification inspection	Personnels do not have health certificates or relevant certificates for pile pitching and pulling.
M10	Personnel management	Including on-site real-time monitoring, personnel management in the pile fixing room, personnel qualification inspection, operation communication and command, emergency response training, operation disclosure, and personnel technical training.	X20	Unclear communication and command of operation	The lifting personnel did not maintain effective contact information with the responsible on-site personnel.
M11	Risk assessment before operation	Including scheme evaluation and vessel evaluation.	X21	No operation disclosure	No operation disclosure was conducted on the plan for pile pitching and pulling, risk evaluation, and the condition of the operating vessels to ensure information symmetry.
M12	Emergency response training	Including the existence of emergency plans, insufficient emergency resources, and poor configuration of life-saving devices.	X22	Personnel technical training	No training was provided to the personnel on pile pitching and pulling, resulting in negligence or inaccurate operation by the personnel.
X1	Excessive wind force	The excessive wind exceeds the wind resistance level of the vessel and the limit of the restoring force arm that can be adjusted in actual operation, resulting in the inability of the vessel to maintain stability and the risk of overturning.	X23	No emergency plans	Failure to develop a safety emergency plan in advance, including how to respond to piercing risks and personnel safety and evacuation plans, such as closely monitoring weather forecasts and the progress of vessel lifting operations to ensure sufficient time to implement emergency plans before adverse weather conditions occur.
X2	Intense waves	Lifting operations should not be carried out in the condition of intense waves.	X24	Poor emergency rescue equipment	The operating vessel is not equipped with life-saving equipment to ensure the safety of the personnel in case of risks.
X3	Unknown submarine geological conditions	Leakage or sudden changes in geological exploration work have not been identified, and the existence of weak strata has not been discovered. Blindly pile pitching poses the risk of piercing.	X25	Inadequate evaluation of homework plan	A suitable and compliant operation plan has not been developed based on the site conditions for the entire process of pre-operation preparation, pile leg lowering, vessel lifting, and pile pulling.
X4	Underwater obstacles	Deformation and sliding of pile shoes caused by the large slope of seabed reefs or strata.	X26	Inadequate assessment of operational vessels	The necessary data such as deck size, configuration, deck height, deck strength, pile leg size, and pile shoes of the platform have not been evaluated.
X5	Changes in water depth	Large changes in tides affect the draft of vessels, resulting in insufficient buoyancy to support them, and waves have a significant impact on the stability of vessels.	X27	Poor physiological condition of personnel	Failure to confirm the physical condition of the personnel before operation resulted in operating in unfavorable conditions such as fatigue and drunkenness.
X6	Vessel instability	Vessel is equipped with stabilizing devices such as anti-roll fins and hydraulic controls or not.	X28	Insufficient professional skills of personnel	The main personnel did not obtain the certificate of pile pitching and pulling or did not meet the technical standards, resulting in the failure of the operation.
X7	Failure of watertight facilities	Defects in the watertight equipment prevent the cabin from maintaining a watertight state.	X29	Lack of standardized education	Not operating in accordance with the vessel operation manual specifications.
X8	Failure of pile driving system	The vessel system was equipped with a failed pile driving system and cannot be used normally. For vessels without a pile driving system, the pile pulling operation cannot reduce the formation friction resistance through high-pressure water.	X30	Poor psychological state of personnel	The employees experience persistent negative emotions, mental distress, or impaired cognitive functioning.
X9	Failure of underwater monitoring system	The platform includes an underwater monitoring system that monitors the real-time erosion of sediment near the pile shoes to prevent the occurrence of erosion and emptying.	X31	Not wearing protective equipment	The safety protection equipment of the personnel was not worn neatly.

provides a detailed list of all events and their meanings that may lead to accidents involving pile pitching and pulling on the offshore jack-up platform. The Fault Tree structure of the failed pile pitching and pulling is shown in Figure 2.

3.2. Mapping Fault Tree to Bayesian Model

The process of transferring information from a Fault Tree to a Bayesian Network entails mapping the event nodes from the Fault Tree to those in the Bayesian Network and establishing directed arc connections that reflect the logical relationships between events [38]. Initially, the nodes in the Bayesian Network and the variables they represent are identified using Table 3. Subsequently, possible states for each node are defined. For instance, the node labeled “excessive wind force (X1)” can be assigned states of “yes” or “no,” indicating whether the wind force surpasses the vessel’s wind resistance level. For other nodes, states can similarly be defined based on real-world conditions. Next, the dependency relationships among the nodes are established. In Bayesian Networks, this is often represented as a directed acyclic graph (DAG), where the node at the arrow’s tail influences the node at the arrow’s head. For example, the “water environmental risk (M5)” might depend on several nodes, including “excessive wind (X1)” and “intense waves (X2).” Finally, a conditional probability table is assigned to each node to outline its conditional probability distribution given the states of its parent nodes. The parent nodes are derived using the fuzzy set theory, as the occurrence of any parent node necessitates the occurrence of its child node, thus employing an OR gate representation. The Bayesian model is depicted in Figure 3.

4. Quantitative Risk Analysis Based on Bayesian Networks

Due to insufficient and ambiguous information in the risk evaluation of pile pitching and pulling on the offshore jack-up platform, this study evaluated the possibility of basic event failure by calculating the probability of node failure based on expert opinions [39]. Based on the established risk index system for pile pitching and pulling on the offshore jack-up platform, this study invited ten experts who have been engaged in offshore wind power construction operations for many years to conduct the scoring and evaluation. These ten experts are all senior engineers in the field of pile pitching and pulling on the offshore jack-up platform, including project managers and safety supervisors with profound practical experience. Detailed information on their academic background (including degree), professional title, work experience, age, and weight, calculated according to Formulas (4) and (5), are shown in

Table 4. Information on the experts.

Expert No.	Degree	Professional Title	Work Experience	Age	Weight
Expert 1	Bachelor	Senior Engineer	15	37	0.095588
Expert 2	Master	Engineer	7	33	0.073529
Expert 3	Bachelor	Professorate Senior Engineer	23	45	0.117647
Expert 4	Bachelor	Professorate Senior Engineer	38	63	0.125000
Expert 5	Bachelor	Professorate Senior Engineer	23	56	0.125000
Expert 6	Bachelor	Senior Engineer	17	39	0.095588
Expert 7	Bachelor	Engineer	12	36	0.080882
Expert 8	Master	Senior Engineer	13	39	0.088235
Expert 9	Bachelor	Engineer	21	43	0.102941
Expert 10	Doctor	Senior Engineer	6	36	0.095588

.

According to the probability of occurrence of basic events, the occurrence level of each event is divided into five levels: very high (VH, red), high (H, orange), medium (M, yellow), low (L, blue), and very low (VL, green) [32]. Based on the knowledge and experience of 10 experts, the degree of occurrence of basic events affecting pile pitching and pulling on the offshore jack-up platform is judged, and the judgment results are shown in

Table 5. Expert judgment of basic events.

Basic Events	Expert 1	Expert 2	Expert 3	Expert 4	Expert 5	Expert 6	Expert 7	Expert 8	Expert 9	Expert 10
X1	H	H	M	M	M	H	H	M	M	M
X2	H	VH	M	H	H	M	H	M	H	M
X3	VH	H	M	H	M	VH	H	H	M	M
X4	L	VL	L	L	L	VL	VL	L	L	L
X5	H	H	VH	M	H	H	M	H	M	H
X6	M	L	L	M	M	M	L	L	M	L
X7	L	L	L	M	L	M	L	L	M	L
X8	L	L	VL	L	VL	L	L	L	L	L
X9	L	L	L	VL	L	L	L	VL	L	VL
X10	L	L	L	L	VL	L	L	L	L	VL
X11	M	M	M	H	M	H	M	M	M	M
X12	L	L	L	L	L	L	L	L	M	L
X13	M	M	L	M	L	M	M	M	L	M
X14	L	VL	L	L	L	L	L	L	L	L
X15	M	M	L	L	L	M	L	L	L	L
X16	L	L	L	M	L	L	L	L	M	M
X17	H	VH	H	M	H	M	M	M	M	H
X18	L	L	L	L	M	M	L	L	M	L
X19	L	L	M	L	L	L	L	L	M	M
X20	L	M	L	L	L	L	L	L	L	M
X21	L	L	VL	L	L	L	L	L	VL	L
X22	L	M	L	M	L	L	L	L	L	M
X23	L	L	M	L	L	M	L	M	L	L
X24	VL	VL	L	L	L	L	L	L	L	L
X25	L	M	L	L	L	L	L	M	L	M
X26	L	VL	L	L	L	L	L	L	L	L
X27	L	L	L	M	M	L	L	L	L	L
X28	M	H	M	H	VH	M	M	M	M	H
X29	L	L	M	L	L	L	L	L	M	L
X30	L	M	L	M	M	L	L	M	M	L
X31	M	M	M	L	L	L	L	L	M	M

.

During expert evaluation, regarding the fuzzy evaluation of system risk factors, this article quantified the subjective evaluation by constructing a fuzzy evaluation set. To obtain accurate probability parameters, Formula (6) was used to weigh and aggregate the opinions of multiple experts, forming a fuzzy probability distribution of group consensus. Subsequently, Formulas (7) and (8) were used to perform defuzzification calculations and convert linguistic variables into numerical probabilities. After the above processing flow, the prior probability distribution of the root node in the Bayesian Network was finally determined, as shown in Figure 4, and the specific parameter values are shown in

Table 6. Fuzzy possibility and prior probability of basic events.

Events	Aggregated Fuzzy Numbers				X	FPr
X1	0.404348	0.586957	0.586957	0.769565	0.586957	0.008981
X2	0.496377	0.662319	0.670290	0.828261	0.663679	0.014639
X3	0.507246	0.669565	0.688406	0.831884	0.672878	0.015514
X4	0.074638	0.186594	0.211957	0.349275	0.207372	0.000253
X5	0.531884	0.691304	0.702899	0.850725	0.693305	0.017650
X6	0.207246	0.384058	0.384058	0.560870	0.384058	0.002026
X7	0.163768	0.329710	0.329710	0.495652	0.329710	0.001216
X8	0.076087	0.190217	0.214130	0.352174	0.209827	0.000263
X9	0.068841	0.172101	0.203261	0.337681	0.197528	0.000213
X10	0.078261	0.195652	0.217391	0.356522	0.213506	0.000279
X11	0.365217	0.554348	0.554348	0.743478	0.554348	0.007252
X12	0.120290	0.275362	0.275362	0.430435	0.275362	0.000666
X13	0.231884	0.414855	0.414855	0.597826	0.414855	0.002627
X14	0.092029	0.230072	0.238043	0.384058	0.236680	0.000398
X15	0.155072	0.318841	0.318841	0.482609	0.318841	0.001088
X16	0.163768	0.329710	0.329710	0.495652	0.329710	0.001216
X17	0.468116	0.638768	0.646739	0.809420	0.640127	0.012616
X18	0.163768	0.329710	0.329710	0.495652	0.329710	0.001216
X19	0.162319	0.327899	0.327899	0.493478	0.327899	0.001194
X20	0.134783	0.293478	0.293478	0.452174	0.293478	0.000824
X21	0.078261	0.195652	0.217391	0.356522	0.213506	0.000279
X22	0.159420	0.324275	0.324275	0.489130	0.324275	0.001151
X23	0.160870	0.326087	0.326087	0.491304	0.326087	0.001172
X24	0.082609	0.206522	0.223913	0.365217	0.220846	0.000314
X25	0.153623	0.317029	0.317029	0.480435	0.317029	0.001067
X26	0.092029	0.230072	0.238043	0.384058	0.236680	0.000398
X27	0.149275	0.311594	0.311594	0.473913	0.311594	0.001007
X28	0.450725	0.623551	0.635870	0.796377	0.625675	0.011511
X29	0.143478	0.304348	0.304348	0.465217	0.304348	0.000931
X30	0.204348	0.380435	0.380435	0.556522	0.380435	0.001962
X31	0.197101	0.371377	0.371377	0.545652	0.371377	0.001810

.

4.1. Information Transmission from Fault Tree to Bayesian Network

GeNle Academic4.1 was used to implement the model transformation; the prior probability of each node in the Bayesian Network was calculated through expert evaluation to form a complete Bayesian Network model [40]. The BN structure is shown in Figure 5, where state0 represents no occurrence, state1 represents occurrence, and most state1 non-zero values are displayed as zero in the figure, due to the display accuracy of the software. The specific values are shown in Table 6.

As shown in Figure 6, in terms of primary indicators, the environmental factor state1 has a probability of 0.525511, indicating a high susceptibility to environmental risk factors. In terms of secondary indicators, factors such as excessive wind force (X1), intense waves (X2), unknown seabed geological conditions (X3), changes in water depth (X5), ineffective on-site supervision (X17), insufficient professional skills of personnel (X28), and the failure of remote sensing equipment (X11) all have a significant impact on the risk level of pile pitching and pulling on the jack-up platform. Therefore, environmental factors have the greatest impact on pile pitching and pulling on the jack-up platform in terms of uncontrollability, suddenness, and variability. They can directly threaten the safety and efficiency of operations, thereby serving as leading factors for accidents. For example, excessive wind force may cause platform overturning, such as the “Fujing 001” accident. The intense waves exacerbate the risk of damage to the platform structure, such as the Bohai No.2 accident. Unclear seabed geological conditions can lead to piercing accidents, such as accidents caused by complex geology in the South China Sea. Changes in water depth affect the stability of operations, such as the “Z” platform flooding accident. Therefore, it is necessary to closely monitor environmental changes and develop emergency plans to ensure operational safety during pile pitching and pulling.

4.2. Probability Update and Sensitivity Analysis

In the application scenario of the offshore jack-up platform (BN), a crucial function is to correct the prior probability based on the latest obtained information, and the probability value obtained from this process is called the posterior probability. The update operation of probability largely relies on an in-depth understanding of key events or outcomes. Specifically, this method starts with the prior probability of the basic event and then sets the occurrence probability of the target node to 100%. On this basis, the posterior probabilities of each node in the network can be further calculated. By comparing and analyzing the difference between prior probability and posterior probability, variables that have a significant impact on triggering unexpected top events can be identified. Figure 6 shows the posterior probability values obtained through reverse analysis, assuming that the top event has already occurred, as shown in Figure 7.

In the established Bayesian Network model, the prior and posterior probabilities obtained are assigned to each node, and the relative rate of variation (ROV) between nodes is derived and quantitatively evaluated, namely, the sensitivity analysis. In sensitivity analysis, the sensitivity of each basic event to the safety status of pile pitching and pulling on the offshore jack-up platform can be effectively evaluated and key driving factors can be identified by accurately comparing the prior probability and posterior probability distributions of basic events and combining the probability difference analysis. Specifically, the larger the probability difference, the higher the sensitivity of the basic event. Posterior probabilities consistently exceed prior values, as shown in Figure 8.

The specific quantification formula is shown in Equation (7): the ROV is obtained by calculating the prior probability and posterior probability of pile pitching and pulling on the offshore jack-up platform, as shown in Figure 9.

$$ROV(X_{\mathrm{i}})={\displaystyle \frac{\pi (X_{\mathrm{i}})-\theta (X_{\mathrm{i}})}{\theta (X_{\mathrm{i}})}}$$

(7)

ROV is the relative change value of Xi, where Xi is the root event, π (Xi) is the posterior probability, and θ (Xi) is the prior probability.

Based on Bayesian Network model analysis, risk factors with significant posterior probability growth contribute significantly to system failure in pile pitching and pulling on the offshore jack-up platform. Through sensitivity analysis, the following key risk points were identified: underwater obstacles (X4), failure of the protective device of the lifting system (X12), unclear pre-pressure inspection before operation (X15), inadequate personnel technical training (X22), low vessel stability (X6), unclear seabed geological conditions (X3), and inadequate on-site supervision (X17).

These key nodes have a multi-level dependency relationship with basic risks with high prior probabilities (such as changes in water depth, unknown seabed geological conditions, intense waves, and ineffective on-site monitoring), forming a transmission chain of “environmental trigger-node failure-chain reaction”, ultimately leading to the entire system accident through the Domino effect. For example, changes in water depth (X5) can indirectly exacerbate the risks posed by underwater obstacles (X4). At the same time, changes in water depth have proposed more stringent requirements for the lifting system equipment (X12), leading to an increased probability of equipment failure. In addition, the unknown geological conditions of the seabed (X3) are a common cause of the risks of underwater obstacles (X4), unclear preloading (X15), and vessel stability (X6). These risk factors are interrelated and affect each other, collectively constituting the potential risks of offshore jack-up platforms. The intense waves (X2) directly exacerbate the risks of vessels and equipment, forming a dual failure chain of “environment equipment”, leading to the instability of vessel stability (X6) and the failure of lifting systems (X12). The insufficient professional skills of personnel (X28) are associated with the failure of the lifting system (X12), unclear preloading (X15), and monitoring failure (X17). Insufficient personnel capabilities are the root cause of operational errors and regulatory failures, forming a “human equipment” risk loop. Excessive wind force (X1) is associated with the stability of the vessel (X6) and the failure of the lifting system (X12), posing a direct threat to the safety of the vessel and equipment. Remote sensing equipment failure (X11) leads to unclear underwater obstacles (X4) and unknown geological conditions (X3). Remote sensing equipment failure is a risk in the information acquisition layer, directly weakening the ability to control the environment and equipment and amplifying other risks. The above risk conduction paths reveal the underlying mechanism by which fundamental risk factors at the prior probability level, such as changes in water depth, unknown seabed geological conditions, and intense waves, trigger the failure of critical nodes, ultimately leading to accidents at the system level.

5. Discussion

This study conducted a systematic analysis on the safety risks of pile pitching and pulling on an offshore wind power jack-up platform, using a comprehensive analysis method of FTA and FBN, and revealed key risk factors and their interaction mechanisms. The research results indicate that environmental risks, such as changes in water depth, excessive wind force, intense waves, and unknown seabed geological conditions, have the most significant impact on the operation of pile pitching and pulling, which is highly consistent with accident cases caused by sudden environmental changes in actual engineering. For example, the direct cause of the overturning accident of the “Shengping 001” platform in 2021 was the coupling effect of increased wind power and complex geological conditions. In addition, equipment failures such as pile washing-out system failures, underwater monitoring system failures, and management deficiencies such as a lack of emergency plans and inadequate on-site monitoring have also been identified as high-risk nodes, which ultimately lead to system level accidents through multi-level conduction paths.

Compared to traditional static risk assessment methods, the dynamic probability model in this study more accurately quantifies the evolution of risk factors within uncertain environments. A sensitivity analysis reveals a high ROV for basic events such as the failure of the pile washing-out system (X8) and the underwater monitoring system (X9), indicating a strong sensitivity to triggering top events. In the context of pile pitching and pulling on offshore wind power jack-up platforms, the failure of the pile washing-out system can lead to severe hazards, including operational interruptions, platform damage, and personnel casualties. For instance, during the operation of the “Ocean Independence” jack-up drilling platform in the Gulf of Mexico, a series of severe accidents occurred due to pile failure, resulting in substantial losses. Operational interruptions not only delay the construction timeline but also increase the safety risks in harsh sea conditions. Platform damage may cause structural deformation or collapse, leading to significant economic losses. The risk of casualties is heightened due to the difficulty in emergency evacuation.

During pile pitching and pulling operations on the offshore jack-up platform, the failure of underwater monitoring systems has led to multiple serious accidents. In 1979, the Bohai No.2 platform overturned due to monitoring failure, resulting in 72 fatalities. In 2006, the anchor chain of the South China Sea Victory broke during a typhoon, causing production to be suspended for nearly a year. In 2020, the Nantong offshore wind power platform was submerged during pile pulling, resulting in the loss of over CNY 187 million. Oil leakage from Norway’s PETROJARL11 platform, caused by an anchor chain breakage, resulted in the compensation of USD 230 million. These accidents are typically caused by sensor failures, interruptions in data transmission, or human–machine misjudgments, leading to platform damage, casualties, and significant economic losses.

To address these risks, a comprehensive prevention and control system should be established from three dimensions: technology, management, and personnel, so as to reduce the probability of risk conduction and the severity of consequences. For example

(1)

High-precision underwater terrain and obstacle detection (such as multi-beam sonar, side-scan sonar) should be conducted before the operation. Known obstacles should be removed or marked, pile positions should be adjusted to avoid risk areas, if necessary, real-time monitoring systems should be installed, and changes in pile leg insertion resistance during the operation should be dynamically monitored.

(2)

The sensitivity and reliability of safety devices should be regularly verified, mandatory maintenance cycles should be established, redundant design (such as dual-channel sensors) should be adopted to ensure that single-point failures do not affect overall safety. Functional testing should be conducted before operation, and system status should be monitored in real time during operation.

(3)

Pre-pressure standards should be established based on the seabed geological report and dynamically adjusted in combination with on-site testing. Pressure sensors should be installed to monitor the pre-pressure values in real time and provide feedback to the control system. Operators need to receive professional training on pre-pressure parameter calculation and adjustment.

(4)

A dual-track training system of “theory+practice” should be implemented, covering equipment principles, operating procedures, and emergency plans, with regular retraining and assessment to ensure that personnel qualifications and skills continue to meet standards. Simulator training should be introduced to enhance operational proficiency under complex working conditions.

(5)

Before operation, stability checks should be conducted on vessels, the distribution of ballast water should be optimized, and inclinometers and anemometers should be installed to monitor the vessel’s attitude and external loads in real-time. Operations in adverse sea conditions should be suspended, and personnel should be evacuated to a safe area if necessary.

(6)

Detailed geological surveys (such as drilling sampling and in situ testing) should be conducted before operation, a geological model should be established, the design parameters of the pile legs should be dynamically adjusted based on historical data and empirical formulas, and pile shoe optimization designs (such as increasing the area and friction surface) should be adopted for complex geological conditions.

(7)

A dedicated safety supervision position should be established, on-site supervision should be implemented throughout the entire process, an AI video surveillance system should be introduced to automatically identify violations and issue alarms, a “dual post confirmation” mechanism should be established, and two or more personnel should be required to review key operations.

6. Conclusions

This study employs the fuzzy set theory to evaluate the probability of basic risk factors associated with complete data loss. Based on the established Fuzzy Bayesian Network model for pile pitching and pulling construction operations of the self- elevating platform in offshore wind power, the probability of the occurrence of basic risk causes calculated by the fuzzy set theory is taken as the input. The forward reasoning capability of the Fuzzy Bayesian Network is harnessed to conduct a reliability analysis of the construction operations, predict the completion probability of each operation step, and assess the occurrence probability of operation influencing factors and the consequences resulting from operation failures. Based on the Bayesian theory, the dynamic failure and accident probabilities of offshore wind power engineering construction are estimated, the risk situation of the construction operations in offshore wind power engineering is analyzed, diagnostic reasoning is performed, and the posterior probability of basic risk factors under the occurrence of evidence is obtained. Through an analysis of the changes between the prior probability and the posterior probability, the basic risk factors that exert a significant influence on the failure of construction operations or the occurrence of consequences in offshore wind power engineering construction can be identified. This study conducted a comprehensive risk evaluation and performed a quantitative cause analysis for pile pitching and pulling events on offshore wind power jack-up platforms by integrating FTA with FBN. Initially, an FTA model was developed based on expert opinions, accident reports, and industry information to identify risk factors and decompose top events into basic events. Subsequently, the FTA events were mapped onto Bayesian Network nodes, and the fuzzy set theory was employed to describe event states and determine prior probabilities. Furthermore, an FBN was constructed to evaluate the likelihood of events based on expert judgment. After fuzzification processing, the scores were quantified, leading to the derivation of fuzzy probability distributions for basic events. Finally, probability updates and sensitivity analyses were carried out to identify key driving factors, and the risk conduction mechanism was elucidated through impact intensity analysis.

The research findings indicate that the critical risk factors impacting the safety of pile pitching and pulling operations can be categorized into three main categories: environmental risks, equipment malfunctions, and management deficiencies. Environmental risks, including abrupt changes in water depth, extreme weather conditions, and uncertain seabed geology, may directly trigger platform instability or equipment failure. Equipment malfunctions, such as failures in the pile washing-out system or underwater monitoring system, may directly lead to operation interruption or accident escalation. Management deficiencies, such as inadequate emergency plans and insufficient on-site supervision, can amplify the consequences of risks. Further analysis reveals a multi-level dependency relationship among these risk factors, forming a chain-like conduction pathway of “environmental triggers—node failures—chain reactions.” For instance, changes in water depth not only heighten the likelihood of collisions with submerged obstacles but also increase the workload on the lifting system, thereby elevating its probability of failure. A sensitivity analysis reveals that failures in the pile washing-out system and monitoring system are highly sensitive events that significantly contribute to top-event occurrences (major accidents), representing a critical link in risk prevention and control.

Based on the aforementioned research findings, this study proposes several improvement measures. These include enhancing risk identification and assessment, optimizing operational processes, improving equipment maintenance and repair, boosting personnel skills and training, upgrading emergency response mechanisms, and promoting technological innovation and research and development. The aim is to reduce the risk level of pile installation and extraction at the source and to enhance the overall safety performance of offshore wind power projects. In summary, this study expands the theoretical and methodological framework for risk assessment and offers effective practical guidance for achieving precise risk prevention and control during pile installation and extraction on offshore wind power jack-up platforms. Looking ahead, as the offshore wind power industry continues to evolve, the research and application of risk assessments and prevention technologies should be further intensified. The safety and reliability of pile installation and extraction should be continuously enhanced to provide robust support for the long-term stable operation of offshore wind power projects. Concurrently, the methodology presented in this study can offer valuable references and insights for security management and risk prevention in other similar fields.