Ship collision risk analysis method and system based on fault tree and Bayesian network

Abstract

The invention provides a ship collision risk analysis method and system based on a fault tree and a Bayesian network, and relates to the technical field of ship safety. The method comprises the following steps: determining that main risk factors are human factors, ship factors, environment factors and management factors by analyzing ship collision accident cases; taking ship collision as a top event, performing deductive reasoning according to occurrence logic, and constructing a fault tree structure; the mapping relation between the fault tree and the Bayesian network is utilized, the Bayesian network of ship collision is established, and basic events, middle events and top events correspond to root nodes, middle nodes and leaf nodes respectively; the prior probability of a root node and the conditional probability of an intermediate node are determined through statistical analysis of main risk factors, and a Bayesian network model is formed; based on the model, a collision risk rate is calculated through forward reasoning, key risk factors are diagnosed through backward reasoning, and a main risk path causing collision is identified. According to the scheme, the complex relation between the risk factors can be accurately and objectively subjected to deep analysis and quantitative description.

Description

Technical Field

[0001] This application relates to the technical field of ship safety. Specifically, it relates to a method and system for ship collision risk analysis based on fault trees and Bayesian networks. Background Technique

[0002] Ship collision risk analysis refers to the process of identifying, evaluating potential collision accidents that may occur during ship navigation and their potential consequences through systematic methods, and then proposing preventive measures. Currently, ship collision risk analysis mainly relies on traditional statistical methods and empirical judgments. These methods usually rely on historical accident data, and through statistical analysis, the correlation and influence degree between various risk factors and ship collision accidents are obtained. However, these methods have many limitations in practical applications.

[0003] On the one hand, traditional statistical methods often can only provide the macroscopic correlation between risk factors and accidents, and it is difficult to deeply reveal the interaction mechanism between various risk factors and their specific influence paths on ship collision accidents. On the other hand, although empirical judgments can reflect the knowledge and experience of experts to a certain extent, they often lack objectivity and accuracy and are easily affected by personal subjective factors. Therefore, there is an urgent need for a solution that can accurately and objectively conduct in-depth analysis and quantitative description of the complex relationships between risk factors to address the shortcomings existing in the above-mentioned prior art. Summary of the Invention

[0004] The purpose of this application is to provide a method and system for ship collision risk analysis based on fault trees and Bayesian networks, which can accurately and objectively conduct in-depth analysis and quantitative description of the complex relationships between risk factors.

[0005] This application is implemented as follows:

[0006] In a first aspect, the present application provides a method for analyzing ship collision risks based on a fault tree and a Bayesian network, including the following steps: Analyze the risk factors of ship collision accidents based on the collected ship collision accident cases to determine the main risk factors, where the main risk factors include human factors, ship factors, environmental factors, and management factors; Based on the determined main risk factors, determine the ship collision accident as the top event, and conduct top-down deductive reasoning according to the occurrence logic of the ship collision accident to establish the fault tree structure of the ship collision accident; According to the mapping relationship between the fault tree structure and the Bayesian network, construct the Bayesian network structure of the ship collision accident, where the basic events, intermediate events, and top event of the fault tree structure are respectively mapped to the root nodes, intermediate nodes, and leaf nodes of the Bayesian network structure; Determine the prior probability of the root nodes based on the statistical analysis of the determined main risk factors, and determine the conditional probability of the intermediate nodes according to the fault tree structure, and obtain the corresponding Bayesian network model based on the Bayesian network structure; Conduct forward reasoning according to the Bayesian network model to calculate the risk rate of the ship collision accident, and conduct backward reasoning according to the Bayesian network model to diagnose the key risk factors of the ship collision accident and identify the main risk paths leading to the ship collision accident.

[0007] In a second aspect, the present application provides a system for analyzing ship collision risks based on a fault tree and a Bayesian network, which includes: A first unit for analyzing the risk factors of ship collision accidents based on the collected ship collision accident cases to determine the main risk factors, where the main risk factors include human factors, ship factors, environmental factors, and management factors; A second unit for determining the ship collision accident as the top event based on the determined main risk factors and conducting top-down deductive reasoning according to the occurrence logic of the ship collision accident to establish the fault tree structure of the ship collision accident; A third unit for constructing the Bayesian network structure of the ship collision accident according to the mapping relationship between the fault tree structure and the Bayesian network, where the basic events, intermediate events, and top event of the fault tree structure are respectively mapped to the root nodes, intermediate nodes, and leaf nodes of the Bayesian network structure; A fourth unit for determining the prior probability of the root nodes based on the statistical analysis of the determined main risk factors, and determining the conditional probability of the intermediate nodes according to the fault tree structure, and obtaining the corresponding Bayesian network model based on the Bayesian network structure; A fifth unit for conducting forward reasoning according to the Bayesian network model to calculate the risk rate of the ship collision accident, and conducting backward reasoning according to the Bayesian network model to diagnose the key risk factors of the ship collision accident and identify the main risk paths leading to the ship collision accident.

[0008] Compared with the prior art, the present application has at least the following advantages or beneficial effects:

[0009] This application proposes a method for analyzing ship collision risks based on fault trees and Bayesian networks. By combining fault tree analysis and Bayesian networks, it can comprehensively and systematically evaluate the risks of ship collision accidents. Fault tree analysis provides a clear logical framework, while Bayesian networks provide powerful probabilistic reasoning capabilities. The combination of the two can significantly improve the accuracy of risk assessment. Among them, through backward reasoning, it is possible to diagnose the key risk factors and main risk paths leading to ship collision accidents. This provides targeted risk management strategies for shipping companies and regulatory agencies, helping them more effectively prevent and control the occurrence of ship collision accidents. In addition, it can not only provide the results of current risk assessment, but also, due to the flexibility of Bayesian networks, allow model updates and optimizations based on new data and situations. This helps shipping companies and regulatory agencies to continuously learn and improve, promoting the continuous improvement of shipping safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0011] Figure 1 It is a flowchart of an embodiment of a method for analyzing ship collision risks based on fault trees and Bayesian networks in this application;

[0012] Figure 2 It is a schematic diagram of the fault tree structure based on human-ship-environment-management in an embodiment of this application;

[0013] Figure 3 It is a schematic diagram of the subtree structure of improper crew behavior events in an embodiment of this application;

[0014] Figure 4 It is a schematic diagram of the subtree structure of poor ship conditions in an embodiment of this application;

[0015] Figure 5 It is a schematic diagram of the subtree structure of poor environment in an embodiment of this application;

[0016] Figure 6 It is a schematic diagram of the subtree structure of improper management faults in an embodiment of this application;

[0017] Figure 7 It is a schematic diagram of the fault tree of ship collision accidents in an embodiment of this application;

[0018] Figure 8 It is a schematic diagram of the Bayesian network structure of ship collision accidents in an embodiment of this application;

[0019] Figure 9 Schematic diagram of the Bayesian network model for ship collision accidents based on fault tree in an embodiment of the present application;

[0020] Figure 10 Schematic diagram of the Bayesian network model for diagnosis of key risk factors in an embodiment of the present application;

[0021] Figure 11 Bayesian network model for identifying main risk paths in an embodiment of the present application. Detailed implementation manners

[0022] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Usually, the components of the embodiments of the present application described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations.

[0023] The following will describe in detail some embodiments of the present application with reference to the accompanying drawings. Without conflict, the various embodiments and the various features in the embodiments below can be combined with each other.

[0024] To facilitate understanding of the technical solutions provided by the present application, some terms in the present application will be introduced first.

[0025] 1. Fault tree analysis method

[0026] Fault tree analysis method (FTA) is a top-down system failure analysis method. Based on accident data and accident occurrence mechanisms, a fault tree is established for combined qualitative and quantitative analysis. The fault tree analysis method can identify the risk factors of each system, analyze the logical relationships between the factors, and represent the logical relationships in a graphical deduction manner. The events in the fault tree include basic events, intermediate events, and top events, and the relationships between various events are described by logic gates.

[0027] When calculating the probability of the top event occurring, first use the Boolean algebra simplification method to calculate the minimum cut sets, and represent the fault tree equivalent diagram with the minimum cut sets, and then use the equivalent diagram to deduce the probability of the top event occurring.

[0028] For a fault tree with k minimum cut sets, the probability of the top event occurring can be expressed as:

[0029]

[0030] In the formula: i is the ordinal number of the bottom event; q i is the minimum cut set; x i is the i-th bottom event; xi ∈ k j The i-th basic event belongs to the j-th minimal cut set; j, S are the ordinals of the minimal cut sets; k is the number of minimal cut sets; x i ∈ k j ∪ k s The i-th basic event x i either belongs to the j-th minimal cut set or belongs to the S-th minimal cut set; 1 ≤ j < s ≤ k is the value range of j, S.

[0031] 2. Bayesian Network

[0032] A Bayesian Network (BN) is a probability model based on Bayes' formula and represented graphically, suitable for solving problems of uncertainty and incompleteness, and has been widely used in risk assessment research. It consists of nodes, arcs, and a Conditional Probability Table (CPT). Nodes represent variables, arcs represent the causal relationships between nodes, and the CPT quantifies the influence degree of parent nodes on their child nodes.

[0033] The theoretical basis for Bayesian network analysis is Bayes' formula. Given new evidence A, the posterior probability can be deduced from the prior probability using Bayes' formula, as shown in the following equation.

[0034]

[0035] In the formula, P(X i |A) is the posterior probability of node X i under the condition that node A occurs, P(X i ) represents the prior probability, represents the conditional probability of node A under the condition that node X i occurs, and P(A) is the total probability.

[0036] In the BN model, the state distribution of a node depends on its parent nodes, that is, when the state of the parent nodes changes, it will have a certain impact on the child nodes. The joint probability distribution of the nodes is:

[0037]

[0038] In the formula, Pa(X i ) represents the set of parent nodes of node X i .

[0039] Example 1

[0040] In the field of ship collision risk analysis, although traditional statistical methods and empirical judgments are widely used, their limitations are becoming increasingly prominent. Among them, traditional statistical methods mainly rely on historical accident data, which can provide the macroscopic correlation between risk factors and accidents, but it is difficult to reveal the interaction mechanism and specific influence path between various risk factors. At the same time, although empirical judgment can reflect the knowledge and experience of experts, it is easily affected by personal subjective factors and lacks objectivity and accuracy. Therefore, the existing technology has obvious deficiencies in the depth and breadth of ship collision risk analysis and cannot provide strong guarantee for shipping safety.

[0041] To overcome these shortcomings of the existing technology, the embodiments of the present application provide a ship collision risk analysis method based on fault tree and Bayesian network. The method aims to deeply analyze and quantitatively describe the complex relationships between various risk factors in a systematic and objective manner, so as to accurately evaluate the ship collision risk and provide a scientific basis for shipping safety.

[0042] Please refer to Figure 1 , the ship collision risk analysis method based on fault tree and Bayesian network includes the following steps:

[0043] Step S101: Analyze the risk factors of ship collision accidents according to the collected ship collision accident cases, and determine the main risk factors, where the main risk factors include human factors, ship factors, environmental factors and management factors.

[0044] It should be noted that by collecting and analyzing ship collision accident cases, the main risk factors affecting ship collision can be comprehensively and systematically identified. This step provides a solid foundation for subsequent fault tree construction and Bayesian network modeling.

[0045] Step S102: According to the determined main risk factors, determine the ship collision accident as the top event, and conduct top-down deductive reasoning according to the occurrence logic of the ship collision accident to establish the fault tree structure of the ship collision accident.

[0046] It should be noted that after determining the main risk factors, step S102 uses fault tree analysis (FTA), takes the ship collision accident as the top event, and conducts top-down deductive reasoning according to the logical order of the accident occurrence to construct a fault tree structure including basic events (i.e., risk factors), intermediate events and top events. This step reveals the logical relationship and action path between risk factors. That is, through the top-down deductive reasoning of fault tree analysis, the logical relationship and action level between various risk factors can be clearly shown, so as to construct the fault tree structure of the ship collision accident. This structure not only helps to understand the logical order of the accident occurrence, but also provides a direct basis for the mapping of the Bayesian network.

[0047] Step S103: Construct a Bayesian network structure for ship collision accidents according to the mapping relationship between the fault tree structure and the Bayesian network. Among them, the basic events, intermediate events, and top event of the fault tree structure are respectively mapped to the root nodes, intermediate nodes, and leaf nodes of the Bayesian network structure. By mapping the fault tree structure to the Bayesian network structure, the transformation from qualitative analysis to quantitative analysis is realized. In the Bayesian network, the basic events, intermediate events, and top event respectively correspond to the root nodes, intermediate nodes, and leaf nodes, which enables the complex relationships between risk factors to be quantitatively described.

[0048] Step S104: Determine the prior probability of the root nodes based on the statistical analysis of the identified main risk factors, and determine the conditional probability of the intermediate nodes according to the fault tree structure, and obtain the corresponding Bayesian network model based on the Bayesian network structure.

[0049] Exemplarily, in the process of establishing a Bayesian network model based on the fault tree structure, since the mapping nodes of the model are not in one-to-one correspondence, it is necessary to optimize the model according to the actual situation. The mapping from the fault tree structure to the Bayesian network model includes two aspects: qualitative processing and quantitative processing.

[0050] (1) Qualitative processing is mainly the mapping of the Bayesian network structure. In terms of network nodes, the basic events, intermediate events, and top event in the fault tree structure are respectively equivalent to the respective nodes in the Bayesian network structure. Among them, the basic events of the fault tree structure correspond to the root nodes of the Bayesian network structure. For the same basic event that appears multiple times, they are merged into one root node in the Bayesian network. In terms of logical relationships, the causal relationships between the nodes in the Bayesian network structure correspond one-to-one with the logical relationships of the corresponding events in the fault tree structure and are connected by directed arcs.

[0051] (2) In quantitative processing, corresponding to the fault tree structure, the states of the network nodes are divided, and the probability distribution of the basic events in the fault tree structure corresponds to the prior probability of the root nodes of the Bayesian network. For the logical gates and intermediate events in the fault tree structure, equivalent CPTs are added to the corresponding nodes in the Bayesian network for description.

[0052] Step S105: Perform forward inference according to the Bayesian network model to calculate the risk rate of ship collision accidents, that is, the probability of the accident occurring under the given risk factors, and perform backward inference according to the Bayesian network model to diagnose the key risk factors of ship collision accidents and identify the main risk paths leading to ship collision accidents.

[0053] In the above steps S101 - S105, by combining fault tree analysis and Bayesian networks, the risks of ship collision accidents can be comprehensively and systematically evaluated. Fault tree analysis provides a clear logical framework, while Bayesian networks offer powerful probabilistic reasoning capabilities. The combination of the two can significantly improve the accuracy of risk assessment. Among them, through backward reasoning, it is possible to diagnose the key risk factors and main risk paths that lead to ship collision accidents. This provides targeted risk management strategies for shipping enterprises and regulatory agencies, helping them more effectively prevent and control the occurrence of ship collision accidents. In addition, it can not only obtain the results of the current risk assessment, but also, due to the flexibility of Bayesian networks, allow for model updates and optimizations based on new data and situations. This helps shipping enterprises and regulatory agencies to continuously learn and improve, promoting the continuous enhancement of shipping safety.

[0054] In summary, the above embodiments provide a new, more comprehensive and systematic method for ship collision risk analysis by combining fault tree analysis and Bayesian networks. This method not only improves the accuracy of risk assessment, but also enhances the pertinence of risk management and promotes the continuous improvement of shipping safety.

[0055] Based on the foregoing solution, in some implementation manners of the present application, the human factors include human factors before the accident and human factors during the accident. Among them, the human factors before the accident include physical and mental state, theoretical level, and practical experience, and the human factors during the accident include operation errors and operation violations; the ship factors include ship emergency failures, on - board equipment, ship tonnage, and ship seaworthiness; the environmental factors include natural environment and navigation environment; the management factors include the external management of ships by the water traffic safety competent department and the internal management of the ship - owning enterprise.

[0056] In the above implementation manners, by further refining and defining the human factors, ship factors, environmental factors, and management factors, the present application can more accurately identify the main risk factors leading to ship collisions, thereby improving the precision of risk assessment. And through this detailed factor division, it helps shipping enterprises and regulatory agencies to more clearly understand the sources of risks, so as to formulate more targeted risk management measures. For example, for crew members with poor physical and mental states, health monitoring and psychological counseling can be strengthened; for crew members who frequently make operation errors, training and assessment can be strengthened. That is, the above implementation manners improve the precision of risk assessment and the pertinence of risk management through the detailed division and definition of human factors, ship factors, environmental factors, and management factors, and promote the continuous improvement of shipping safety.

[0057] Specifically, through summarizing the previous research on risk factors of ship collision accidents and analyzing ship collision accident cases, it can be finally determined that the risk factors of ship collision accidents are four main risk factors: human, ship, environment, and management. And a comprehensive analysis is carried out on what specific risk factors there are under each main risk factor. Among them, the human factor, ship factor, environmental factor, and management factor are described in detail as follows:

[0058] (1) Human factor

[0059] As the main influencing factor leading to water accidents, the human factor also plays the most important role in collision accidents. According to relevant causation theories, the human factor in collision accidents can be divided into two categories. One is the indirect factor generated based on the accumulation of the crew's own conditions before the accident; the other is the relevant factor directly causing the accident due to the improper behavior of the crew during the accident.

[0060] 1) Human factors before the accident

[0061] Physical and mental state: Here it refers to the impact of the crew's physical state and mental state on navigation work, such as adverse states like fatigue and drunkenness.

[0062] Theoretical level: Here it refers to whether the crew participating in navigation work has received sufficient training and education, whether they have mastered the basic knowledge of ship operation and water area navigation, and whether they have learned and understood relevant navigation regulations knowledge.

[0063] Practical experience: Here it refers to the crew's familiarity with navigation conditions such as relevant water areas, ship equipment status, and transported goods, as well as their various abilities to operate the ship. Being able to accurately receive and process relevant information and prevent accidents is an embodiment of the good navigation experience of excellent crew members.

[0064] 2) Human factors during the accident

[0065] Operation errors: Here it refers to the phenomenon of improper operation behavior during the accident caused by factors such as insufficient preparation, insufficient understanding of navigation conditions, and lack of personal ability. Specific manifestations in collision accidents include improper route selection, improper ship operation, and emergency operation errors.

[0066] Operation violations: Here it refers to the situation where the crew does not comply with relevant maritime laws and regulations during navigation. Specific manifestations in collision accidents include improper lookout, risky navigation, failure to give way as required, and failure to use a safe speed.

[0067] (2) Ship factor

[0068] In a collision accident, the overall condition of the ship is an important influencing factor. When the ship is in a poor condition, it is prone to restricted ship maneuverability, so that the ship cannot normally execute the maneuvering orders of the commanding personnel when encountering sudden accidents, resulting in serious consequences. The condition and performance of ship equipment, as well as the relevant characteristics and attributes of the hull structure, can be reflected by the following factors.

[0069] 1) Ship emergency failures

[0070] It refers to the system failures such as sudden main engine failure, propeller failure, and steering gear failure of ship equipment when the ship faces a dangerous situation. It has a certain degree of contingency. Although the daily maintenance and inspection work has been in place, it will still occur during navigation due to unforeseen factors.

[0071] 2) Onboard equipment

[0072] Onboard equipment mainly includes intercoms, VHF radios, life-saving and fire-fighting equipment, etc. Marine traffic accident investigations show that the lack of onboard equipment or the failure of onboard equipment is one of the causes of quite a number of accidents. Therefore, the equipment configuration and usage of onboard equipment are also closely related to navigation safety.

[0073] 3) Ship tonnage

[0074] The gross tonnage of the ship reflects the size of the ship, and the degree of danger of ships of different tonnages is also different.

[0075] 4) Ship seaworthiness

[0076] Ship seaworthiness mainly refers to whether the ship meets the basic conditions for navigation, which is generally measured from three aspects: one is whether the ship's certificates are complete; the second is whether the crew meets the minimum safe manning requirements; the third is whether the ship's structure or the loaded cargo is normal. If abnormal, it will lead to hull damage and uneven loading. As a basic element of ship safety navigation, ship seaworthiness is the key to ensuring navigation safety and requires key attention.

[0077] (3) Environmental factors

[0078] In terms of environmental factors, the variability of the natural environment often causes changes in navigation conditions, and the quality of the traffic environment directly affects the ship's maneuvering space. When ship operators cannot respond to environmental changes in a timely manner, the ship will be in a dangerous situation. For collision accidents, environmental impact factors cannot be ignored.

[0079] 1) Natural environment

[0080] ① Wind and waves

[0081] When sailing at sea, the forces acting on the ship's water surface mainly come from wind and waves. Excessive wind and waves increase the forces on the ship, affecting ship maneuverability and having an impact on navigation safety.

[0082] ② Flow velocity

[0083] In a collision accident, the flow velocity is one of the more important influencing factors. When a ship approaches relevant facilities, it is often necessary to maintain an appropriate speed. However, when a ship is sailing in waters with a relatively fast flow velocity, its maneuverability decreases, and it cannot achieve the effect of sailing at an appropriate speed, which poses a great test to the safety of the ship.

[0084] ③ Visibility

[0085] Poor visibility caused by weather conditions such as heavy fog restricts the vision of the on-duty crew, making it difficult to accurately judge the positional relationship between the ship and relevant facilities. In this state, relevant accidents are extremely likely to occur.

[0086] 2) Navigable environment

[0087] ① Channel depth

[0088] During the navigation of a ship, when the water depth is too shallow, the shallow water effect will occur, and the motion characteristics of the ship will change. If the crew is not familiar with the relevant waters, serious consequences are likely to occur.

[0089] ② Channel width

[0090] The narrowness of the channel directly affects the size of the ship's maneuvering space. When sailing in a narrow water area, problems such as the bank effect are likely to occur, which poses a great test to the crew's maneuvering ability. In the field of collision accidents, accidents caused by limited water area width are not uncommon.

[0091] ③ Aids to navigation

[0092] Aids to navigation refer to navigation aids that are usually installed in the channel where necessary to help ships navigate and avoid dangerous obstacles in order to achieve safe and convenient ship navigation. Whether there are problems with the setting of aids to navigation directly affects the decisions of the ship's crew and the traffic efficiency of the water area. Properly set aids to navigation are an important guarantee for the safe navigation of ships.

[0093] ④ Ship traffic volume

[0094] Ship traffic volume refers to the number of ships passing through a certain point in the water area per unit time, which reflects the density of ship navigation in this water area and is one of the important risk factors affecting water navigation safety.

[0095] (4) Management factors

[0096] The safe navigation of ships on the water is inseparable from good management measures. Without timely and correct constraints from managers, water traffic accidents are likely to occur. Therefore, management factors are usually regarded as the main risk factors that can cause water traffic accidents. Management factors are generally divided into two aspects: external management of ships by water traffic safety authorities and internal management of the ship's owning companies.

[0097] 1) External management

[0098] ① Laws and regulations

[0099] A sound legal system is the basis for ensuring water transportation safety. Maritime regulations must ensure that all aspects of navigation safety are covered to serve as the code of conduct for seafarers.

[0100] ②Supervision and management

[0101] It refers to whether the water safety regulatory department can fully perform its regulatory responsibilities for ship navigation, including comprehensive supervision of navigation, ships, crew members, etc., and promptly discover and correct inappropriate behaviors during navigation or even behaviors that violate relevant regulations.

[0102] 2) Internal management

[0103] ①Safety Management System

[0104] It refers to whether the shipping company or operator has established a safety management system and whether the established safety management system can achieve various management requirements.

[0105] ② Adequate management

[0106] It refers to whether the enterprise can carry out adequate safety management work in accordance with the established safety management system, including organizing various safety trainings and improving the crew's operating capabilities.

[0107] like Figure 2 As shown, based on the above scheme, in some implementations of the present application, the ship collision accident is determined as the top event, and the deductive reasoning from top to bottom is performed according to the occurrence logic of the ship collision accident, including: determining the ship collision accident as the top event, and performing the deductive reasoning from top to bottom according to the occurrence logic of the ship collision accident. After the reasoning from the first layer to the second layer, a fault tree structure based on man-ship-environment-management is obtained, wherein the subtree structure includes improper personnel behavior, poor ship status, poor environment and improper management. Qualitative analysis is performed on the subtree, and the fault tree subtree is constructed in a top-down deduction order.

[0108] Among them, by constructing a fault tree structure based on human-vessel-environment-management, this application can comprehensively cover the main risk factors leading to ship collision accidents, ensuring the comprehensiveness and accuracy of risk analysis. Moreover, through qualitative analysis of subtrees, this application can identify the key factors and main paths leading to ship collision accidents, providing specific risk management measures and operation guidelines for shipping enterprises and regulatory agencies.

[0109] Based on the foregoing solution, in some implementation manners of this application, as Figure 3 shown, in the subtree structure of improper human behavior events, the logical relationship between layers is represented by an AND gate. Among them, during the deduction process, improper human behavior is divided into two intermediate events: problems occurring before the accident (P1) and problems occurring during the accident (P2); before the accident, three basic events are obtained: poor physical and mental state (P11), insufficient theoretical level (P12), and insufficient practical experience (P13); during the accident stage, two intermediate events are obtained: operation error (P21) and operation violation (P22), which respectively include three basic events: improper route selection (P211), improper ship handling (P212), and improper emergency handling (P213), and four basic events: improper lookout (P221), risky navigation (P222), failure to give way as required (P223), and failure to adopt a safe speed (P224).

[0110] As Figure 4 shown, in the subtree structure of poor ship condition events, the logical relationship between layers is connected by an OR gate. Among them, during the deduction process, poor ship condition includes ship emergency failure (S1), on-board equipment failure (S2), excessive ship tonnage (S3), and unseaworthiness of the ship (S4) events. Among them, unseaworthiness of the ship is an intermediate event, and three basic events are deduced: incomplete ship certificates (S41), failure to meet the minimum safe manning (S42), and abnormal ship structure or cargo (S43).

[0111] As Figure 5 shown, in the subtree structure of poor environment events, the logical relationship between layers is connected by an OR gate. Among them, during the deduction process, two intermediate events are first deduced: poor natural environment (E1) and poor navigation environment (E2); in terms of the natural environment, a set of basic events composed of excessive wind and waves (E11), excessive water flow velocity (E12), and poor visibility (E13) is deduced; in terms of the navigation environment, according to the logical relationship between events, four basic events are obtained: too shallow water depth of the waterway (E21), insufficient width of the waterway (E22), problems with navigation aids (E23), and large ship traffic volume (E24).

[0112] As Figure 6As shown in the figure, in the subtree structure of mismanagement events, the logical relationships between layers are connected by AND gates. Among them, in the deduction process, it is first divided into two intermediate events: external mismanagement (M1) by regulatory departments such as maritime agencies and internal mismanagement (M2) by the ship - owning company, and these two are further deduced; for the mismanagement of regulatory departments, it is deduced into two basic events: imperfect laws and regulations (M11) and insufficient supervision (M12); for the mismanagement of the ship company, it is deduced into two basic events: imperfect safety management system (M21) and insufficient management (M22).

[0113] Integrating all subtree structures, the fault tree of ship collision accidents is as Figure 7 shown, where relevant events are represented by their codes. Through the above - mentioned analysis, the event descriptions and numbers of the ship collision fault tree are shown in Table 1.

[0114] Table 1 Fault tree event descriptions and numbers

[0115]

[0116]

[0117] Exemplarily, based on the above implementation method, the establishment and analysis of the Bayesian network model for ship collision accidents are as follows:

[0118] (1) Bayesian network structure construction

[0119] There is a mutual conversion relationship between the structure of the fault tree and the Bayesian network structure. According to the mapping relationship between the fault tree structure and the Bayesian network, it is converted into a Bayesian network structure. The basic events of the ship collision accident fault tree can be mapped to the root nodes of the Bayesian network, and then advanced layer by layer according to relevant logic, and finally the conversion of the Bayesian network structure of the ship collision accident is realized. The Figure 7 ship collision fault tree is converted into a Bayesian network structure as Figure 8 shown, where 27 basic events and 13 intermediate events are respectively mapped to root nodes and intermediate nodes, and the top event "ship collision accident" is mapped to a leaf node.

[0120] (2) Bayesian network model analysis

[0121] 1) Model establishment

[0122] 1.1) Calculation of prior probabilities of root nodes

[0123] Exemplarily, 20 ship collision accident reports from the Yangtze River Waterway Bureau from 2014 to 2024 were collected for statistical analysis to determine the statistical probability and fuzzy numbers of the occurrence of each root node event, and then the final prior probability of the occurrence of the root node event was obtained using Equation (9) in the following text. The calculation results are shown in Table 2.

[0124] Table 2 Prior Probabilities of the Occurrence of Root Node Events

[0125] root node prior probability root node prior probability root node prior probability P11 0.11 P224 0.30 E13 0.11 P12 0.30 S1 0.11 E21 0.11 P13 0.50 S2 0.11 E22 0.11 P211 0.13 S3 0.11 E23 0.11 P212 0.13 S41 0.11 E24 0.11 P213 0.70 S42 0.30 M11 0.11 P221 0.70 S43 0.11 M12 0.50 P222 0.30 E11 0.11 M21 0.30 P223 0.30 E12 0.11 M22 0.70

[0126] 1.2) Establishment of Bayesian Network Model

[0127] Based on the calculation of the prior probabilities of the root nodes of ship collision accidents, the conditional probabilities of each intermediate node were determined using a fault tree, and forward reasoning was performed using GeNIe software to obtain the probabilities of all intermediate nodes and the occurrence of ship collision accidents. The established Bayesian network model of ship collision accidents is shown in Figure 9 as shown. The risk rate of ship collision accidents is 0.97.

[0128] 2) Diagnosis of Key Risk Factors

[0129] The diagnosis of key risk factors is a bottom-up backward reasoning of the Bayesian network model. When the occurrence probability of the target node is set to 100%, the key factors leading to the accident can be identified. Taking the occurrence of ship collision accidents as a hypothetical condition, the posterior probabilities of each basic event are inferred. The posterior probability reflects the degree of possibility that the root node leads to the occurrence of ship collision accidents. The larger the posterior probability, the greater the possibility that the node becomes the cause of ship collision accidents. Setting the probability of the leaf node of ship collision accidents to 100%, through Figure 9 the established Bayesian model for backward reasoning, the reasoning results are shown in Table 3 and Figure 10 as shown. Table 3 and Figure 10 show that the root nodes with higher posterior probabilities are P221, P213, M22, P13, and M12 in sequence. The above factors respectively represent improper lookout, improper emergency response measures, insufficient management, lack of practical experience, and insufficient supervision, indicating that these factors have a greater impact on ship collision accidents. Targeted prevention and control of these factors can effectively reduce the risk rate of ship collision accidents.

[0130] Table 3 Posterior Probability Distribution of Root Nodes

[0131]

[0132] 3) Identification of Main Risk Paths

[0133] The main risk paths leading to ship collision accidents can be identified by backward reasoning using GeNIe software. The specific steps are as follows:

[0134] 3.1) Set the status of the root node "ship collision accident" to occurred.

[0135] 3.2) Calculate the posterior probabilities of each node. Starting from the "ship collision accident" node, sequentially search backward for the parent node with the largest posterior probability to establish the most important risk path of the ship collision accident. Finally, the maximum risk path leading to the ship collision accident is: improper lookout - operation violation - problems during the accident - improper personnel behavior - ship collision accident, as shown by the bold arrow markings in Figure 11 as shown.

[0136] Based on the foregoing solution, in some implementation manners of the present application, when establishing the fault tree structure of the ship collision accident, the relationships between various events are described using logic gates including AND gates and OR gates.

[0137] It should be noted that in the fault tree, an AND gate means that the output event will occur only when all input events occur simultaneously. This reflects the situation where certain factors must exist simultaneously to cause a ship collision. Contrary to the AND gate, an OR gate means that the output event will occur as long as one input event occurs. This reflects the situation where certain factors alone are sufficient to cause a ship collision. By using AND gates and OR gates to describe the relationships between various events, the complex logical chain of the occurrence of the ship collision accident can be accurately reflected.

[0138] Based on the foregoing solution, in some implementation manners of the present application, determining the prior probability of the root node according to the statistical analysis of the determined main risk factors includes: for basic events with historical data, using probability statistical analysis to obtain the prior probability of the occurrence of the basic event in each ship collision event; for basic events lacking historical data, using fuzzy theory knowledge to obtain the probability of the event occurrence, converting the statistically obtained prior probability into a fuzzy number through a fuzzy membership function, and then using the centroid method for defuzzification to convert the fuzzy number into a fuzzy probability.

[0139] It should be noted that for basic events with historical data, the above implementation manner uses the probability statistical analysis method. By collecting, organizing, and analyzing relevant historical data, the prior probability of the occurrence of the basic event in each ship collision event is calculated. This method is based on the statistical results of a large amount of historical data and can relatively accurately reflect the objective law of the occurrence of the basic event.

[0140] For basic events lacking historical data, the above implementation uses fuzzy theory knowledge to obtain the probability of event occurrence. Among them, fuzzy theory is an effective tool for dealing with uncertainty problems. It allows the probability of event occurrence to have a certain degree of fuzziness, that is, it is not a deterministic 0 or 1, but a value between 0 and 1. In practical applications, probability assessment of basic events lacking historical data can be carried out through methods such as expert scoring and questionnaire surveys, combined with fuzzy theory knowledge.

[0141] To convert a fuzzy number into a fuzzy probability, the above implementation uses a fuzzy membership function to describe the mapping relationship between the fuzzy number and the fuzzy probability. A fuzzy membership function is a function that describes the membership degree of elements in a fuzzy set. It can convert a fuzzy number into a fuzzy probability value between 0 and 1. In the above implementation, an appropriate fuzzy membership function can be selected according to the distribution characteristics of the fuzzy number to convert the statistically prior probability into a fuzzy number, and then defuzzification is carried out through the centroid method to convert the fuzzy number into a fuzzy probability. By converting the fuzzy number into a definite fuzzy probability value, it can be used in subsequent fault tree analysis and risk assessment.

[0142] In summary, by adopting probability statistical analysis methods and fuzzy theory knowledge, this application can obtain the prior probability of basic events more accurately, thereby improving the accuracy of risk assessment. Especially for basic events lacking historical data, the application of fuzzy theory knowledge can make up for the deficiency of data, making the risk assessment more comprehensive and reliable. And this implementation allows the probability of event occurrence to have a certain degree of fuzziness, which enhances the flexibility of risk management.

[0143] In other words, the risk factors of ship collision are complex and diverse. There are mainly two ways to obtain the prior probability of basic events: ① For basic events with a large amount of historical data, probability statistical analysis can be used to obtain the probability of event occurrence; ② For the lack of data, fuzzy theory knowledge can be used to obtain the probability of event occurrence.

[0144] Based on the analysis of ship collision cases, the fuzzy probability of ship collision can be obtained by integrating the two methods. The steps for calculating the prior probability of basic events of ship collision based on historical data and fuzzy theory are as follows:

[0145] (1) Obtain the initial prior probability

[0146] Collect historical events of ship collision and count the prior probability of basic events occurring in each ship collision event.

[0147] (2) Divide the probability levels

[0148] Divide the probabilities of basic events occurring into five levels: "High (H)", "Higher (MH)", "Medium (M)", "Lower (ML)", and "Low (L)".

[0149] (3) Convert the statistically obtained prior probabilities into fuzzy numbers

[0150] Fuzzy numbers include triangular fuzzy numbers, trapezoidal fuzzy numbers, normal fuzzy numbers, etc. Triangular fuzzy numbers and trapezoidal fuzzy numbers are simple to operate and have been widely used. According to the five divided probability intervals (u H , u MH , u M , u ML and u L ), determine the fuzzy membership function formula as:

[0151]

[0152]

[0153] Conduct statistical analysis on the collected samples of ship collision accidents and calculate the initial prior probabilities of each root node occurring. According to the fuzzy membership function, obtain the fuzzy numbers of each root node event occurring.

[0154] (4) Convert the fuzzy numbers into fuzzy probabilities

[0155] Adopt the centroid method for defuzzification to convert the fuzzy numbers into fuzzy probabilities. The calculation formula is shown in Equation (9).

[0156]

[0157] In the formula, f, f2, f3, f4 are the values in the trapezoidal fuzzy number in sequence; P * is the fuzzy probability converted from the fuzzy number, that is, the prior probability of the root node.

[0158] Based on the foregoing solution, in some implementation manners of the present application, the mapping relationship includes qualitative processing and quantitative processing, where the qualitative processing is the mapping of the Bayesian network structure, and the quantitative processing includes dividing the states of network nodes and corresponding the probability distribution of basic events in the fault tree to the prior probabilities of the root nodes of the Bayesian network.

[0159] Among them, in the process of establishing the BN model based on the FTA structure, since the mapping nodes of the model are not in one-to-one correspondence, it is necessary to optimize the model according to the actual situation. The mapping from FTA to BN includes two aspects: qualitative processing and quantitative processing:

[0160] (1) The qualitative processing is mainly the mapping of the BN structure. In terms of network nodes, the basic events, intermediate events, and top events in FTA are respectively equivalent to the various nodes in the BN model. Among them, the basic events in FTA correspond to the root nodes of BN. For the same basic event that appears multiple times, they are merged into one root node in BN. In terms of logical relationships, the causal relationships between nodes in BN correspond one-to-one with the logical relationships of the corresponding events in FTA and are connected by directed arcs.

[0161] (2) In the quantitative processing, corresponding to FTA, the states of network nodes are divided, and the probability distribution of the basic events in FTA corresponds to the prior probability of the root nodes in BN. For the logical gates and intermediate events in FTA, equivalent CPTs are added to the corresponding nodes in BN for description.

[0162] Based on the foregoing solution, in some implementation manners of the present application, when performing backward reasoning according to the Bayesian network model, it includes: taking the occurrence of a ship collision accident as a hypothesis condition, and inferring the posterior probability of each basic event, where the posterior probability reflects the degree of possibility that the root node causes the occurrence of the ship collision accident; and identifying the main risk path leading to the ship collision accident through backward reasoning.

[0163] Among them, in the process of backward reasoning, first set the occurrence of the ship collision accident as a hypothesis condition, that is, the state of the target node (or called the top event) is known. Then, using the reasoning algorithm of the Bayesian network, according to the set hypothesis condition and the known prior probability and conditional probability, calculate the posterior probability of each basic event (i.e., the root node). The posterior probability reflects the degree of possibility that each basic event causes the accident under the condition that the ship collision accident has occurred. Next, by analyzing the calculated posterior probability, identify the main risk path leading to the ship collision accident. The main risk path refers to those basic events with higher posterior probabilities and their interdependencies that have an important impact on the occurrence of the ship collision accident.

[0164] Exemplarily, an exact reasoning algorithm (such as variable elimination method, clique tree propagation method, etc.) or an approximate reasoning algorithm (such as belief propagation algorithm, Monte Carlo simulation method, etc.) can be used to perform backward reasoning.

[0165] Based on the foregoing solution, in some implementation manners of the present application, when identifying the main risk path leading to the ship collision accident, it includes: starting from the ship collision accident node, sequentially searching backward for the parent node with the largest posterior probability to establish the main risk path of the ship collision accident.

[0166] It should be noted that by searching backward from the ship collision accident node to find the parent node with the highest posterior probability, the key factors leading to the accident can be deeply explored, improving the depth of risk identification. The identified main risk paths provide a scientific basis for the formulation of risk management and preventive measures, making risk management more targeted and effective. Moreover, by continuously monitoring and analyzing the changes in the main risk paths, new risk factors and potential safety hazards can be detected in a timely manner, promoting the continuous improvement and enhancement of safety management. Also, when a ship collision accident occurs, the problem can be quickly located based on the identified main risk paths, and corresponding emergency measures can be taken to improve the efficiency and accuracy of emergency response.

[0167] That is, the above implementation method constructs the main risk path by searching backward from the ship collision accident node to find the parent node with the highest posterior probability, realizing an in-depth analysis of the root cause of the ship collision accident and an accurate identification of the key risk factors.

[0168] Embodiment 2

[0169] The embodiment of the present application provides a ship collision risk analysis system based on a fault tree and a Bayesian network, which includes:

[0170] A first unit for analyzing the risk factors of ship collision accidents based on the collected ship collision accident cases to determine the main risk factors, where the main risk factors include human factors, ship factors, environmental factors, and management factors; a second unit for determining the ship collision accident as the top event according to the determined main risk factors, and performing top-down deductive reasoning according to the occurrence logic of the ship collision accident to establish the fault tree structure of the ship collision accident; a third unit for constructing the Bayesian network structure of the ship collision accident according to the mapping relationship between the fault tree structure and the Bayesian network, where the basic events, intermediate events, and top events of the fault tree structure are respectively mapped to the root nodes, intermediate nodes, and leaf nodes of the Bayesian network structure; a fourth unit for determining the prior probability of the root nodes according to the statistical analysis of the determined main risk factors, and determining the conditional probability of the intermediate nodes according to the fault tree structure, and obtaining the corresponding Bayesian network model based on the Bayesian network structure; a fifth unit for performing forward reasoning according to the Bayesian network model to calculate the risk rate of the ship collision accident, and performing backward reasoning according to the Bayesian network model to diagnose the key risk factors of the ship collision accident and identify the main risk paths leading to the ship collision accident.

[0171] For the specific implementation process of the above system, please refer to a ship collision risk analysis method based on a fault tree and a Bayesian network provided in Embodiment 1, which will not be elaborated here.

[0172] For those skilled in the art, it is obvious that the present application is not limited to the details of the above-described exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present application is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present application. Any reference signs in the claims should not be construed as limiting the claims concerned.

Claims

1. A method for analyzing the risk of ship collision based on fault tree and Bayesian network, characterized in that, It includes the following steps: Analyze the risk factors of ship collision accidents based on the collected ship collision accident cases, and determine the main risk factors, which include human factors, ship factors, environmental factors, and management factors; Based on the determined main risk factors, determine the ship collision accident as the top event, and conduct top-down deductive reasoning according to the occurrence logic of ship collision accidents to establish the fault tree structure of ship collision accidents; According to the mapping relationship between the fault tree structure and the Bayesian network, construct the Bayesian network structure of ship collision accidents, where the basic events, intermediate events, and top events of the fault tree structure are respectively mapped to the root nodes, intermediate nodes, and leaf nodes of the Bayesian network structure; Based on the statistical analysis of the determined main risk factors, determine the prior probabilities of the root nodes, and determine the conditional probabilities of the intermediate nodes according to the fault tree structure, and obtain the corresponding Bayesian network model based on the Bayesian network structure; Conduct forward reasoning according to the Bayesian network model to calculate the risk rate of ship collision accidents, and conduct backward reasoning according to the Bayesian network model to diagnose the key risk factors of ship collision accidents and identify the main risk paths leading to ship collision accidents.

2. The method according to claim 1, wherein The human factors include human factors before the accident and human factors during the accident. Among them, the human factors before the accident include physical and mental state, theoretical level, and practical experience, and the human factors during the accident include operation errors and operation violations; The ship factors include ship emergency failures, on-board equipment, ship tonnage, and ship seaworthiness; The environmental factors include natural environment and navigation environment; The management factors include the external management of ships by the water traffic safety competent department and the internal management of the ship-owning enterprise.

3. The method according to claim 1, wherein Determining the ship collision accident as the top event and conducting top-down deductive reasoning according to the occurrence logic of ship collision accidents includes: Determine the ship collision accident as the top event, and conduct top-down deductive reasoning according to the occurrence logic of ship collision accidents. After reasoning from the first layer to the second layer, obtain the fault tree structure based on human-ship-environment-management, where the subtree structure includes improper human behavior, poor ship condition, poor environment, and improper management; Conduct qualitative analysis on the subtree and construct the fault tree subtree according to the top-down deduction order.

4. The method according to claim 3, wherein In the subtree structure of the improper human behavior event, the logical relationship between layers is represented by an AND gate. Among them, during the deduction process, improper human behavior is divided into two intermediate events: problems occurring before the accident and problems occurring during the accident; before the accident, three bottom events of poor physical and mental state, insufficient theoretical level, and insufficient practical experience are obtained; during the accident stage, two intermediate events of operation error and operation violation are obtained, including three bottom events of improper route selection, improper ship operation, and improper emergency handling, and four bottom events of improper lookout, adventurous navigation, failure to give way as required, and failure to adopt a safe speed respectively; In the subtree structure of ship condition adverse events, the logical relationships between layers are connected by OR gates. Among them, during the deduction process, ship condition adverse includes ship emergency failures, onboard equipment failures, excessive ship tonnage, and unseaworthy ship events. Among them, unseaworthy ship is an intermediate event, and three basic events of incomplete ship certificates, non-compliance with the minimum safe manning, and abnormal ship structure or cargo are deduced; In the subtree structure of environmental adverse events, the logical relationships between layers are connected by OR gates. Among them, during the deduction process, two intermediate events of adverse natural environment and adverse navigation environment are first deduced; in terms of the natural environment, a set of basic events composed of excessive wind and waves, excessive water flow velocity, and poor visibility are deduced; in terms of the navigation environment, according to the logical relationships between events, four basic events of too shallow water depth in the waterway, insufficient waterway width, problems with navigation aids, and heavy ship traffic are obtained; In the subtree structure of improper management events, the logical relationships between layers are connected by AND gates. Among them, during the deduction process, it is first divided into two intermediate events of external improper management by regulatory departments such as maritime agencies and internal improper management of the ship's owner company, and further deductions are made for these two; in terms of improper management by regulatory departments, it is deduced that two basic events of imperfect regulations and insufficient supervision can be deduced; in terms of improper management by the ship company, it is deduced that two basic events of imperfect safety management system and insufficient management are deduced.

5. The method according to claim 1, wherein When establishing the fault tree structure of ship collision accidents, the logical relationships between various events are described using logical gates including AND gates and OR gates.

6. The method according to claim 1, characterized in that The prior probabilities of root nodes are determined according to the statistical analysis of the identified main risk factors, including: For basic events with historical data, probability statistical analysis is used to obtain the prior probabilities of the occurrence of basic events in each ship collision event; For basic events lacking historical data, fuzzy theory knowledge is used to obtain the probability of event occurrence, and the statistical prior probability is transformed into a fuzzy number through a fuzzy membership function, and then the centroid method is used for defuzzification to transform the fuzzy number into a fuzzy probability.

7. The method according to claim 1, wherein The mapping relationship includes qualitative processing and quantitative processing. Among them, qualitative processing is the mapping of the Bayesian network structure, and quantitative processing includes dividing the states of network nodes and corresponding the probability distribution of basic events in the fault tree to the prior probabilities of the root nodes of the Bayesian network.

8. The method according to claim 1, wherein When performing reverse reasoning based on the Bayesian network model, it includes: taking the occurrence of a ship collision accident as a hypothetical condition, inferring the posterior probabilities of each basic event, and this posterior probability reflects the degree of possibility of the root node causing the ship collision accident; and identifying the main risk path leading to the ship collision accident through reverse reasoning.

9. The method according to claim 1 or 8, characterized in that When identifying the main risk path leading to the ship collision accident, it includes: starting from the ship collision accident node, sequentially searching backward for the parent node with the largest posterior probability to establish the main risk path of the ship collision accident.

10. A ship collision risk analysis system based on fault tree and Bayesian network, characterized in that, Including: The first unit is used to analyze the risk factors of ship collision accidents based on the collected ship collision accident cases, determine the main risk factors, and the main risk factors include human factors, ship factors, environmental factors, and management factors; The second unit is used to determine the ship collision accident as the top event according to the determined main risk factors, and conduct top-down deductive reasoning according to the occurrence logic of the ship collision accident to establish the fault tree structure of the ship collision accident; The third unit is used to construct the Bayesian network structure of the ship collision accident according to the mapping relationship between the fault tree structure and the Bayesian network, wherein the basic events, intermediate events and top events of the fault tree structure are respectively mapped to the root nodes, intermediate nodes and leaf nodes of the Bayesian network structure; The fourth unit is used to determine the prior probability of the root node according to the statistical analysis of the determined main risk factors, and determine the conditional probability of the intermediate node according to the fault tree structure, and obtain the corresponding Bayesian network model based on the Bayesian network structure; The fifth unit is used to conduct forward reasoning according to the Bayesian network model to calculate the risk rate of the ship collision accident, and conduct backward reasoning according to the Bayesian network model to diagnose the key risk factors of the ship collision accident and identify the main risk path leading to the ship collision accident.

Images