Maritime channel security risk assessment method based on multi-state fuzzy Bayesian network

Abstract

The present invention discloses a maritime channel security risk assessment method based on a multi-state fuzzy Bayesian network. The Bayesian model and fuzzy set theory are combined to assess the risks of key nodes. Based on the statistical analysis of historical data, a directed acyclic graph of key nodes is constructed from five aspects: the natural environment, navigation environment, non-traditional security environment, military and political environment, and legal and international environment of the key nodes. The combination of the Bayesian model and fuzzy set theory effectively handles the multi-state problem. The expert survey method based on the confidence index compensates for the cognitive uncertainty problem of the Bayesian model in the parameter learning process by introducing probability intervals, and can more realistically reflect the actual risk situation of the key nodes.

Description

Technical Field

[0001] The present invention relates to the technical field of emergency management systems, and in particular to a maritime channel safety risk assessment method based on a multi-state fuzzy Bayesian network. Background Art

[0002] Key nodes in maritime corridors are crucial for maritime transport, and their safety is crucial for ensuring efficient, stable, and smooth maritime transport for vessels. However, the interplay of various risk factors at these nodes has led to a constant occurrence of various accidents, posing a significant threat to their safety.

[0003] Bayesian networks (BNs) are theoretical models for representing and reasoning uncertain knowledge, and are widely used in risk analysis, assessment, and reliability analysis. Maritime passageways are complex, uncertain systems, with risks at key nodes influenced by multiple factors. Existing quantitative analysis techniques struggle to effectively analyze these risks. MFBNs, based on fuzzy set theory and graph theory, enable fusion analysis of mixed, multi-factor uncertainties. Therefore, a maritime passageway safety risk assessment method based on multi-state fuzzy Bayesian networks has become an urgent challenge. Summary of the Invention

[0004] The technical problem to be solved by the present invention is a maritime channel safety risk assessment method based on a multi-state fuzzy Bayesian network.

[0005] To solve the above technical problems, the present invention provides a technical solution for maritime channel security risk assessment based on a multi-state fuzzy Bayesian network, comprising the following steps:

[0006] Step 1: Identification of risk influencing factors: Collect data statistics on historical cases of maritime channels, analyze risk source factors, identify risk influencing factors, conduct causal relationship analysis on identified risk influencing factors, and establish a risk assessment accident tree model for key nodes of maritime channels;

[0007] Step 2: Construction of MFBN: Construct a directed acyclic graph through the node state description of the risk assessment accident tree model, and construct a multi-state fuzzy conditional probability table that describes the uncertain logical relationship between nodes;

[0008] Step 3: Deductive reasoning based on MFBN: Calculate the probability distribution of risks at key nodes in the maritime channel and diagnose the impact of various risk factors on the safety of key nodes through sensitivity analysis;

[0009] Step 4: Result analysis: Input risk influencing factors in real time to analyze the risk assessment results of key nodes in the maritime channel.

[0010] As an improvement, the historical cases of sea lanes in step one include navigation environment, natural environment data, non-traditional security environment, military and political environment, and legal and international environment data.

[0011] As an improvement, the risk assessment accident tree model includes bottom events, intermediate events and top events.

[0012] As an improvement, the steps of constructing a directed acyclic graph in step 2 are:

[0013] 1) The node risk level of the risk assessment accident tree model is used as the corresponding node state and expressed by fuzzy numbers;

[0014] 2) The bottom event, intermediate event and top event in the risk assessment accident tree model correspond to the root node, intermediate node and leaf node in the MFBN respectively;

[0015] 3) The direction of the directed edge corresponds to the input-output relationship of the logic gate of the risk assessment accident tree model, that is, the input event is the parent node Y, and the output event is the child node T to construct a directed acyclic graph.

[0016] As an improvement, the step of constructing a multi-state fuzzy conditional probability table describing the uncertain logical relationship between nodes in step 2 is:

[0017] 1) Establish neutral judgment criteria ξ and ψ and confidence index δ i ξ is marked with “I, II, III, IV, V…”. The smaller the mark value is, the less credible the judgment result is. ψ is marked with a fuzzy number in the range of 0-1. The smaller the ψ value is, the less credible the judgment result is. The confidence index δ i =ξ i ×ψ i ;

[0018] 2) Divide the possibility P describing the causal relationship between variables into probability intervals i, A, based on the multi-state fuzzy Bayesian network, the risk assessment method for maritime channel security i =[k i ,m i ,k i+1 ],m i represents the average value;

[0019] 3) Collect data and establish a multi-state fuzzy conditional probability table;

[0020] 4) Data processing and defuzzification analysis.

[0021] As an improvement, the data parameters of the multi-state fuzzy conditional probability table include: a parent node Y, a child node T, a neutral judgment criterion ψ, and a possibility P, wherein the parent node Y has two parameters y1 and y2.

[0022] As an improvement, the defuzzification analysis includes the following steps:

[0023] 1) If the result of P(T|y1|y2) corresponding to the Nth neutral judgment criterion ψ is A in , then the fuzzy possibility interval of M neutral judgment criteria ψ is: in represents the fuzzy possibility interval;

[0024] 2) The formula for defuzzifying the fuzzy possibility interval using the α-weighted valuation method is as follows:

[0025] in Indicates the exact value after defuzzification.

[0026] As an improvement, the probability distribution formula for calculating the risk of key nodes in the sea channel in step 3 is:

[0027]

[0028] As an improvement, the sensitivity analysis in step three adopts an analysis method based on information entropy to analyze the degree of influence of factors affecting the safety of maritime channels.

[0029] As an improvement, information entropy is a statistic that describes the degree of discreteness of random variables. When information entropy increases, the uncertainty of the variable also increases. Its calculation formula is as follows: Where is the information entropy of the random variable Y and is the prior probability of Y.

[0030] As an improvement, mutual information can reflect the degree of dependence between two variables, thereby calculating the impact of risk factors on maritime channel safety. The calculation formula is as follows: Where P(x,y) represents the joint probability distribution function of X and Y, and p(x) and p(y) are the marginal probability distribution functions of X and Y, respectively.

[0031] Compared with the existing technology, the advantages of the present invention are: combining the Bayesian model and fuzzy set theory to evaluate the risks of key nodes; on the basis of statistical analysis of historical data, constructing a directed acyclic graph of key nodes from five aspects: the natural environment, navigation environment, non-traditional security environment, military and political environment, and legal and international environment of key nodes; the combination of the Bayesian model and fuzzy set theory effectively handles multi-state problems; the expert survey method based on confidence indicators compensates for the cognitive uncertainty problem of the Bayesian model in the parameter learning process by introducing probability intervals, and can more realistically reflect the actual risk situation of key nodes. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a flow chart of the maritime channel security risk assessment method based on multi-state fuzzy Bayesian network of the present invention.

[0033] Figure 2 This is a statistical chart of historical cases of risk events at key nodes in maritime channels.

[0034] Figure 3 This is the schematic diagram of the correspondence between the logic gates in FT and the CPT in BN.

[0035] Figure 4 It is a directed acyclic graph of risk events at key nodes of maritime channels.

[0036] Figure 5 is a fuzzy possibility interval diagram. DETAILED DESCRIPTION

[0037] The following is a further detailed description of the maritime channel security risk assessment method based on a multi-state fuzzy Bayesian network of the present invention in conjunction with the accompanying drawings.

[0038] Combined with the attached figure, the specific implementation process of the maritime channel security risk assessment method based on multi-state fuzzy Bayesian network is as follows:

[0039] Identification of influencing factors and DAG: Data statistics are collected on historical cases of maritime channels (in this embodiment, historical cases from 2015 to 2020 are counted), and the statistical table is as follows: Figure 2 As shown in the figure, risk source factor analysis and risk influencing factors are carried out. The main factors affecting the risks of key nodes in maritime channels can be summarized into five categories: natural environment, navigation environment, non-traditional security environment, military and political environment, and legal and international environment. According to the analysis results, a safety accident tree model of key nodes in maritime channels is constructed. The direction of the directed edge corresponds to the input-output relationship of the logic gate of the risk assessment accident tree model (FT) to obtain a DAG graph. Among them, the corresponding relationship between the logic gate in FT and the CPT in BN is as follows: Figure 3 As shown, the transformed DAG diagram is as follows Figure 4 As shown;

[0040] Figure 4 The state of each variable in has five security levels, represented by IV, corresponding to five fuzzy numbers. The lower the level, the lower the risk level. The specific division results are shown in Table 1 below. The specific analysis process is as follows:

[0041] The natural environment at a critical node refers to the weather conditions when a ship passes through a critical node, including wind, waves, visibility, and extreme weather conditions. Extreme weather conditions such as tsunamis, storms, and sandstorms increase the probability of risk events occurring at a critical node.

[0042] The navigation environment of a critical node refers to the physical form of the critical node, such as width, depth, and substitutability. The width, depth, and number of substitutable nodes of a critical node will limit the node's traffic capacity, thereby increasing the probability of risk events occurring at the node.

[0043] The non-traditional security environment of key nodes refers to factors that may affect the security of key nodes caused by non-traditional security threats, including piracy and terrorism.

[0044] The military and political environment at key nodes refers to a series of potential impacts related to the military and political environment when a ship is sailing at sea. This includes the number of countries to which key nodes belong, the presence of military bases, and the degree of war risk.

[0045] The legal and international environment at key nodes refers to the legal and international cooperation conditions within the region or coastal countries through which a key node passes, including the domestic and international legal constraints and the number of organizations within the host country. A robust international legal environment and mature organizations can enhance the resilience of sea lanes. The maritime strategies and domestic legal constraints of the coastal countries also impact the safety of the lanes.

[0046] Table 1 The description and discretization of variables

[0047]

[0048]

[0049] To avoid outliers or invalid data, this example surveyed five experts with more than 10 years of work experience and compiled a multi-state fuzzy conditional probability table for leaf node T, as shown in Table 2. Taking P(T = t | y1 = 1, y2 = 4) as an example, the detailed survey results and calculation process are shown in Table 3.

[0050] Table 2 Fuzzy conditional probability table of leaf node T

[0051] Fig.2MFCPT of leaf node T in the DAG

[0052]

[0053] Table 3 Survey and analysis results P(T=t|y1=1,y2=4)

[0054] Fig.3Investigation and analysis results for P(T=t|y1=1,y2=4)

[0055]

[0056]

[0057] The data is defuzzified and analyzed. The specific principles are as follows:

[0058] Assume that the nth expert's estimate of P(T=1|y1=1,y2=1) is A in , then the fuzzy possibility interval result synthesized by M experts is shown in formula (1):

[0059]

[0060] in, Indicates the reliability of the data estimated by the nth expert, δ n represents the confidence index of the nth expert.

[0061] In order to accurately obtain the CPT parameters, it is necessary to Defuzzification is performed and the α-weighted estimation method is applied to defuzzify the fuzzy possibility interval. The specific calculation process is shown in formula (2):

[0062]

[0063] in Indicates the exact value after defuzzification, F α ={x|F(x)≥α} represents the α level set of the set F, f(α) is the weighted assignment function of α, Average(F α ) represents the average value of the α level set, which can be calculated by formula (3), u α and v α Respectively represent the upper and lower bounds of the α level set. Figure 5 As shown, u α and v α It can be calculated by formula (4).

[0064] For example, when f(α)=1, α=5, The calculation result of is shown in formula (5). Similarly, The calculation result of can also be obtained through the above process. In order to meet the normalization conditions, use formula (6) to Perform normalization to obtain the final accurate normalization parameters Similarly, other parameters The MFCPT of each node is finally determined through the above calculation process.

[0065]

[0066] The processed data is input into the MFBN model as initial evidence, and the deductive reasoning of the following formula (7) is implemented using NETICA software;

[0067]

[0068] where P(T=t|X1=x1,X2=x2,...,X n =x n ,Y1=y1,Y2=y2,...,Y m =y m ) represents the conditional probability of leaf node T;

[0069] Where P(X1=x1,X2=x2,...,X n =x n ,Y1=y1,Y2=y2,..., represents the joint probability of the root node and the intermediate nodes, {t1,t2,...,t p} represents the risk level set of leaf nodes, and represents the i-th root node X i The risk level set, Indicates the jth intermediate node Y j A set of risk levels.

[0070] The probability distribution of the leaf nodes in the initial state is obtained as shown in formula (8). The fuzzy numbers are converted into normalized values ​​using formulas (5)-(6), as shown in Table 4 below. The results show that the risk level of the key nodes of the sea channel is medium based on the prior probability. As new evidence is continuously added, the risk warning results will also change accordingly. The states under different scenarios (scenario A, B, C, D) are input into the MFBN model as updated evidence, and the risk levels of the sea channel under different scenarios are analyzed. The results are shown in Table 4. From the output results, it can be seen that the risk level of scenario D is the lowest, and its natural environment, non-traditional security environment, military and political environment, and legal and international environment have all been improved accordingly. The improvement of the military and political environment as well as the legal and international environment in scenario C only reduces the risk of the sea channel to a small extent. The risks of the natural environment and non-traditional security environment are reduced in scenarios A and B respectively, and the risk of the entire sea channel is reduced from medium risk to low risk. By collecting information on internal and external environmental changes through real-time monitoring of the channel environment and inputting the above information into the MFBN model as the latest evidence for real-time deductive reasoning, the risk status of the maritime channel can be dynamically grasped, providing a basis for decision-making for maritime stakeholders.

[0071]

[0072] Table 4 Probability distribution at key nodes of sea lanes of communications under different scenarios

[0073]

[0074] Apply NETICA software and formulas A sensitivity analysis of the BN was conducted, using mutual information to represent the relationships between nodes. The results are shown in Table 5. Overall, the navigation environment has the highest mutual information with key node risks, accounting for 31.7%, followed by the natural environment and the non-traditional security environment. The legal and international environment, as well as the military and political environment, have relatively little impact on key nodes. For the navigation environment, the highest mutual information is with ship traffic at the root node, accounting for 32%. Ship congestion at nodes increases the probability of traffic accidents, leading to increased risk at key nodes. Extreme weather has the highest mutual information with the natural environment. Major accidents such as shipwrecks, collisions, and groundings are generally affected by extreme weather. Piracy has the highest mutual information with the non-traditional security environment and is a significant factor influencing key node risks.

[0075] Table 5 Mutual information between parent node and child node

[0076]

[0077] The present invention and its embodiments are described above. Such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by the above, and does not deviate from the purpose of the present invention, without inventive design, a structure and embodiment similar to the technical solution should fall within the scope of protection of the present invention.

Claims

1. A maritime channel security risk assessment method based on multi-state fuzzy Bayesian networks is characterized by: The following steps are involved: Step 1: Identification of risk influencing factors: Collect data statistics on historical cases of maritime channels, analyze risk source factors, identify risk influencing factors, conduct causal relationship analysis on identified risk influencing factors, and establish a risk assessment accident tree model for key nodes of maritime channels; Step 2: Construction of MFBN: Construct a directed acyclic graph through the node state description of the risk assessment accident tree model, and construct a multi-state fuzzy conditional probability table that describes the uncertain logical relationship between nodes; Step 3: Deductive reasoning based on MFBN: Calculate the probability distribution of risks at key nodes in the maritime channel and diagnose the impact of various risk factors on the safety of key nodes through sensitivity analysis; Step 4: Result analysis: Input risk influencing factors in real time to analyze the risk assessment results of key nodes in the sea channel; The historical cases of sea lanes in step 1 include navigation environment, natural environment data, non-traditional security environment, military and political environment, and legal and international environment data; The risk assessment accident tree model includes bottom events, intermediate events and top events; The steps of constructing a directed acyclic graph in step 2 are: 1) The node risk level of the risk assessment accident tree model is used as the corresponding node state and expressed by fuzzy numbers; 2) The bottom event, intermediate event and top event in the risk assessment accident tree model correspond to the root node, intermediate node and leaf node in the MFBN respectively; 3) The direction of the directed edge corresponds to the input-output relationship of the logic gate of the risk assessment accident tree model, that is, the input event is the parent node Y, and the output event is the child node T to construct a directed acyclic graph; The steps of constructing a multi-state fuzzy conditional probability table describing the uncertain logical relationship between nodes in step 2 are: 1) Establish neutral judgment criteria ξ and ψ and confidence index δ i ξ is marked with "I, II, III, IV, V...", the smaller the mark value, the lower the credibility of the judgment result; ψ is marked with a fuzzy number in the range of 0-1, the smaller the ψ value, the lower the credibility of the judgment result. The confidence index δ i =ξ i ×ψ i ; 2) Divide the possibility P describing the causal relationship between variables into probability intervals i, A, based on the multi-state fuzzy Bayesian network, the risk assessment method for maritime channel security i =[k i ,m i ,k i+1 ],m i represents the average value; 3) Collect data and establish a multi-state fuzzy conditional probability table; 4) Data processing and defuzzification analysis; The multi-state fuzzy conditional probability table data parameters include: parent node Y, child node T, neutral judgment standard ψ and possibility P, where the parent node Y has two parameters y1 and y2; The defuzzification analysis includes the following steps: 1) If the result of P(T|y1|y2) corresponding to the Nth neutral judgment criterion ψ is A in , then the fuzzy possibility interval of M neutral judgment criteria ψ is: in represents the fuzzy possibility interval; 2) The formula for defuzzifying the fuzzy possibility interval using the α-weighted valuation method is as follows: in Indicates the exact value after defuzzification; The probability distribution formula for calculating the risk of key nodes in the sea channel in step 3 is: The sensitivity analysis in step 3 uses an information entropy-based analysis method to analyze the impact of factors affecting the safety of maritime channels; Information entropy is a statistic that describes the degree of discreteness of a random variable. When information entropy increases, the uncertainty of the variable also increases. Its calculation formula is as follows: Where is the information entropy of the random variable Y, is the prior probability of Y; Mutual information can reflect the degree of dependence between two variables, thereby calculating the impact of risk factors on maritime channel safety. The calculation formula is as follows: Where P(x,y) represents the joint probability distribution function of X and Y, and p(x) and p(y) are the marginal probability distribution functions of X and Y, respectively.

Images