Inland unmanned ship static and dynamic coupling risk analysis and evolution prediction method and system

Abstract

The application relates to an inland unmanned ship static-dynamic coupling risk analysis and evolution prediction method and system, the method constructs an algorithm-ship physical entity-navigation environment-shore-based control and management four-dimensional risk index system, and based on historical accident data mapping and N-K model calculation of multi-factor risk coupling value, obtains static coupling strength sorting and key coupling mode; an expert correction coefficient is introduced to calibrate the node failure rate of a dynamic Bayesian network, a multi-time slice dynamic Bayesian network is constructed for forward reasoning, and the cumulative failure probability and dynamic evolution curve of the risk factors and coupling state evolution over time are output. The application can quantize the nonlinear coupling effect of multi-dimensional risks, realize the promotion from static snapshot analysis to dynamic trend prediction, solve the new risk identification and coupling evolution prediction problem of unmanned ships, and provide effective support for risk prevention and control and dynamic safety decision.

Description

Technical Field

[0001] This invention relates to the field of ship navigation risk assessment technology, and in particular to a method and system for static and dynamic coupled risk analysis and evolution prediction of unmanned inland waterway vessels. Background Technology

[0002] With the development of intelligent sensing, automatic control, ship-to-shore communication, and remote control technologies, inland waterway vessels are transitioning from traditional manned operation to unmanned or remotely controlled operation. Compared to open sea waters, inland waterways are characterized by limited navigation space, dense traffic flow, complex bridge and lock areas, and frequent hydrological and meteorological changes. Unmanned inland waterway vessels are more susceptible to the combined influence of various factors, including onboard intelligent functions, the vessel's physical structure, the external navigation environment, and shore-based control and management.

[0003] Existing ship risk assessment methods mostly focus on single risk factors or static causal relationships, such as fault tree analysis, failure mode and effects analysis, and static Bayesian networks. These methods can describe the relationship between some risk factors and accident consequences, but they usually struggle to reveal the nonlinear coupling relationships between multiple risk factors, and they also fail to reflect the time accumulation and dynamic propagation characteristics of risks during continuous navigation.

[0004] For unmanned inland waterway vessels, accidents are not always directly caused by a single factor. More often, they are the result of a coupling effect where factors such as perception anomalies, decision-making biases, physical execution failures, environmental disturbances, and delays in shore-based takeover extricate each other. Therefore, static risk identification alone or dynamic probabilistic reasoning alone is insufficient to fully reflect the risk evolution mechanism. There is an urgent need for a method that can first identify the static coupling strength between risk factors and then characterize the dynamic evolution of risk over time, in order to improve the effectiveness of risk prediction and early warning for unmanned inland waterway vessels. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology by providing a method and system for static and dynamic coupling risk analysis and evolution prediction of unmanned inland waterway vessels, so as to solve or partially solve the problem that traditional risk analysis lacks the ability to quantitatively identify the nonlinear coupling effects of multiple risk dimensions and predict the evolution of risk status over time.

[0006] The objective of this invention can be achieved through the following technical solutions: One aspect of the present invention provides a method for static and dynamic coupling risk analysis and evolution prediction of unmanned inland waterway vessels, comprising the following steps: Construct a risk indicator system for unmanned inland waterway vessels, dividing the risk indicators into primary risk dimensions and secondary risk factors; The system acquires historical accident data of inland waterway vessels, filters and maps the historical accident data, maps the traditional causes of vessel accidents to the risk indicator system, generates a risk factor state matrix, and calculates the marginal probability of each risk factor and the joint probability of the combination of risk factors. Based on the NK model, two-factor, three-factor, and four-factor risk coupling values ​​are calculated, and static coupling strength ranking and key static coupling patterns are obtained according to the risk coupling values. A baseline failure rate is set, and the marginal probabilities of the risk factors are corrected based on a preset expert correction coefficient. The corrected probabilities are then converted into the failure rate of the dynamic Bayesian network nodes. A dynamic Bayesian network model is constructed, and risk factor nodes, risk dimension nodes, and risk coupling nodes are set on continuous time slices. The state transition probability between adjacent time slices is determined based on the failure rate. Based on the dynamic Bayesian network model, dynamic coupling inference is performed to output the cumulative failure probability under different time slices and generate dynamic coupling evolution curves.

[0007] As a preferred technical solution, in the dynamic Bayesian network model, the risk coupling nodes include CP, CE, CS, PE, PS, ES, CPE, CPS, CES, PES, and CPES.

[0008] As a preferred technical solution, the process of correcting the marginal probability of the risk factor based on a preset expert correction coefficient includes: The first The weight of renowned experts is recorded as follows: , and make it for the first The revised score for each risk factor is denoted as: Then the first Expert correction coefficients for each risk factor ; The corrected probability is ; Failure rate of the dynamic Bayesian network nodes ,in This is the system-level baseline failure rate.

[0009] As a preferred technical solution, the state transition probability between adjacent time slices is: For the Risk factor nodes Transition from normal state To failure state probability If the slice was already in an invalid state at a previous time, then ,in The dynamic Bayesian network model employs the first-order Markov assumption, with adjacent time slice intervals.

[0010] As a preferred technical solution, the risk coupling value The calculation formula is: in, Indicates by A combination of risk dimensions Representing the state variables of each risk dimension, The joint probability of the combination of risk dimensions. For the first Marginal probabilities of each risk dimension.

[0011] As a preferred technical solution, the process of obtaining the static coupling strength ranking and key static coupling modes based on the risk coupling value includes the following steps: Calculate the risk coupling values ​​for two-factor, three-factor, and four-factor scenarios, compare the magnitudes of each risk coupling value to obtain a ranking of static coupling strength, and identify the combination with the larger risk coupling value as the key static coupling mode.

[0012] As a preferred technical solution, the risk factor state matrix is ​​composed of the risk factor state vectors of each accident sample. For the first... j The accident and the first i One risk factor; if the risk factor exists, then the state value... x ij = 1, otherwise x ij = 0; The marginal probability is ,in The total number of accident samples. For including the first i Number of accident samples for each risk factor; The combination of risk factors A joint probability ,in To simultaneously include combinations A The number of accident samples for each risk factor in China.

[0013] As a preferred technical solution, the function mapping includes: In traditional ship accidents, improper lookout is mapped to a failure of perception logic; misjudgment or improper collision avoidance decision is mapped to a failure of decision planning; improper maneuvering is mapped to a failure of motion control logic; equipment failure is mapped to risk factors related to the ship's physical entity; meteorological, hydrological, and waterway constraints are mapped to risk factors related to the navigation environment; and insufficient management, supervision, communication, and remote support are mapped to risk factors related to shore-based control and management.

[0014] As a preferred technical solution, the primary risk dimension includes algorithms, ship physical entities, navigation environment, and shore-based control and management; among which, The algorithm dimension includes failures in perception logic, decision planning, motion control logic, and defects in algorithm training data; The physical entity dimension of a ship includes failures of power and actuators, failures of sensing and communication hardware, failures of onboard computing platforms, and failures of automatic loading and stability. The navigation environment dimension includes severe weather interference, complex hydrological interference, and restricted navigation environment; The shore-based control and management dimensions include remote monitoring takeover failures, cybersecurity and management vulnerabilities, and insufficient remote maintenance support.

[0015] As a preferred technical solution, it also includes: perturbing the failure rate of each risk factor node, calculating the change in the posterior probability of the overall risk node and the risk coupling node before and after the perturbation, and obtaining the risk factor sensitivity ranking based on the magnitude of the change.

[0016] Another aspect of the present invention provides a static-dynamic coupling risk analysis and evolution prediction system for inland waterway unmanned vessels, used to implement the aforementioned method for static-dynamic coupling risk analysis and evolution prediction of inland waterway unmanned vessels, the system comprising: The risk indicator system construction module is used to construct a risk indicator system for unmanned inland waterway vessels. The risk indicator system includes four primary risk dimensions: algorithm, vessel physical entity, navigation environment, and shore-based control and management, as well as their respective secondary risk factors. The historical data mapping and probability calculation module is used to acquire historical accident data of inland waterway vessels, filter and functionally map the historical accident data to map the traditional causes of vessel accidents to the risk indicator system, generate a risk factor state matrix, and calculate the marginal probability of each risk factor and the joint probability of the combination of risk factors. The static coupling analysis module is used to encode the risk dimension state of each accident sample, calculate the two-factor, three-factor and four-factor risk coupling values ​​based on the NK model, and obtain the static coupling strength ranking and key static coupling patterns based on the risk coupling values. The failure rate determination module is used to set a baseline failure rate, introduce an expert correction coefficient to correct the marginal probability of the risk factor, and convert the corrected probability into the failure rate of the dynamic Bayesian network node. The Dynamic Bayesian Network Construction Module is used to construct a multi-time-slice dynamic Bayesian network model. In each time slice, risk factor nodes, risk dimension nodes, and risk coupling nodes are set, and the state transition probability between adjacent time slices is determined based on the failure rate. The dynamic coupling inference module is used to perform forward inference based on the dynamic Bayesian network model, output the cumulative failure probability of each node under different time slices, and generate dynamic coupling evolution curves. The sensitivity analysis module is used to perturb the failure rate of each risk factor node, calculate the change in the posterior probability of the overall risk node and risk coupling node before and after the perturbation, and obtain the sensitivity ranking of risk factors based on the change.

[0017] Compared with the prior art, the present invention has at least one of the following beneficial effects: (1) Identification of strongly correlated combinations of risk coupling: Existing methods are difficult to quantify the strength of nonlinear coupling between multiple risk dimensions and cannot determine which dimension combinations are prone to risk amplification effect. Based on the NK model and risk dimension state coding, this invention calculates two-factor, three-factor and four-factor risk coupling values ​​and outputs static coupling strength ranking and key coupling patterns. It can objectively identify strongly correlated combinations of risk coupling from historical accident data, and provide quantitative basis for formulating targeted decoupling control measures and risk prevention priorities.

[0018] (2) High credibility and scenario adaptability: There are few actual accident samples of unmanned ships. Directly using the statistical probability of traditional ship accidents will cause the risk prediction results to deviate from the actual situation due to scenario differences. To this end, the present invention introduces an expert correction coefficient to calibrate the marginal probability of risk factors calculated based on historical data, and combines it with the system benchmark failure rate to convert it into the failure rate of dynamic Bayesian network nodes. This effectively integrates sparse objective data with domain expert experience, and improves the credibility and scenario adaptability of risk probability parameters in the new scenario of unmanned ships.

[0019] (3) Predicting the future growth trend and propagation path of risks: This invention constructs a dynamic Bayesian network model with multiple time slices, uses Markov assumptions and node failure rates for forward reasoning, and outputs the cumulative failure probability and dynamic coupling evolution curve of each node under different time slices. It can simulate the dynamic propagation and accumulation process of risk status in the time dimension, and realize the improvement from static snapshot analysis to dynamic trend prediction, providing support for preventive maintenance and dynamic safety decision-making. Attached Figure Description

[0020] Figure 1 This is a flowchart of the static and dynamic coupling risk analysis and evolution prediction method for unmanned inland waterway vessels in the embodiment; Figure 2 This is a diagram illustrating the risk indicator system in the example. Figure 3 This is a schematic diagram of the NK static coupling analysis process in the embodiment; Figure 4 This is a schematic diagram of the DBN dynamic coupling analysis structure in the embodiment; Figure 5 This is a schematic diagram of the static coupling sorting and dynamic coupling evolution curves in the embodiment; Figure 6 This is a schematic diagram of the static and dynamic coupling risk analysis and evolution prediction system for unmanned inland waterway vessels in the embodiment. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0022] Example 1 To address the problem that existing risk assessment methods cannot simultaneously characterize the static nonlinear coupling relationship and dynamic temporal evolution process of risk factors, this embodiment provides a static and dynamic coupled risk analysis and evolution prediction method for unmanned inland waterway vessels.

[0023] See Figure 1 The method includes the following steps: Step S1: Construct a risk indicator system for unmanned inland waterway vessels, dividing the risk indicators into primary risk dimensions and secondary risk factors.

[0024] like Figure 2 As shown, this embodiment addresses the operation scenario of unmanned inland waterway vessels and constructs a risk indicator system encompassing computation, physical entity, environment, and shore-based control and management. Here, computation corresponds to the algorithm, physical entity to the vessel, environment to the navigation environment, and shore-based control and management, forming the primary risk dimension. Secondary risk factors include: Failures in perception logic, decision planning, motion control logic, and algorithm training data at the algorithm level; Failures of power and actuators, sensing and communication hardware, shipboard computing platform, and automatic loading and stability in the physical dimension of the ship; Severe weather interference, complex hydrological interference, and restricted navigation environment in the navigation environment dimension; The issues include the failure of remote monitoring takeover under shore-based control and management, network security and management vulnerabilities, and insufficient remote maintenance support.

[0025] Figure 2 This table illustrates the hierarchical relationship between primary risk dimensions and secondary risk factors in the aforementioned risk indicator system. To further define the technical meaning and accident identification boundaries of each secondary risk factor, Table 1 explains the specific meaning of each secondary risk factor. This indicator system includes four primary risk dimensions and fourteen secondary risk factors.

[0026] Table 1 Secondary Risk Factors Step S2: Obtain historical accident data of inland waterway vessels, filter and functionally map the historical accident data to generate a risk factor state matrix, and calculate the marginal probability and joint probability of risk factors based on the risk factor state matrix.

[0027] This embodiment acquires historical accident data for inland waterway vessels, filters the accident data, and removes accidents outside of inland waterway scenarios, accidents lacking clear causal chains, and accidents with incomplete liability determinations. For the retained accident samples, a functional mapping method is used to map the causes of traditional vessel accidents to the risk indicator system for unmanned inland waterway vessels.

[0028] Specifically, the causes of accidents related to improper lookout in traditional ships are mapped to perception logic failures; judgment errors or improper collision avoidance decisions are mapped to decision planning failures; improper maneuvering is mapped to motion control logic failures; equipment failures are mapped to risk factors related to the ship's physical entity; meteorological, hydrological, and waterway constraints are mapped to risk factors related to the navigation environment; and insufficient management, supervision, communication, and remote support are mapped to risk factors related to shore-based control and management.

[0029] Establish a risk factor state vector for each accident sample. For the th... j The accident and the first i If a risk factor exists in the accident sample, then let... x ij =1; if the risk factor does not exist, then let x ij =0. The state vectors of all accident samples form the risk factor state matrix.

[0030] Based on the aforementioned risk factor state matrix, calculate the... i Marginal probability of each risk factor P i : in, NThis represents the total number of accident samples. n i Indicates containing the first i Number of accident samples for each risk factor.

[0031] For risk factor combinations A Calculate their joint probability P A : in, n A This indicates that a combination of risk factors is included. A The number of accident samples for each risk factor in China.

[0032] Step S3: Based on the NK model, perform static coupling analysis on different combinations of risk dimensions, calculate the risk coupling values ​​for two-factor, three-factor, and four-factor factors, and output the static coupling strength ranking and key static coupling patterns according to the risk coupling values.

[0033] like Figure 3 As shown, in this embodiment, the algorithm, ship physical entity, navigation environment, and shore-based control and management are denoted as C, P, E, and S, respectively, and a four-dimensional risk state code is applied to each accident sample. For example, state code 1110 indicates that the algorithm, ship physical entity, and navigation environment are all involved in the cause of the accident, while shore-based control and management are not involved in the cause of the accident.

[0034] After completing the state coding, the two-factor, three-factor, and four-factor risk combinations are statistically analyzed to obtain the corresponding marginal probabilities, joint probabilities, and combination probabilities obtained through marginalization. For any risk dimension combination... A Its NK risk coupling value T A Calculate using the following formula: in, A Indicates by m A combination of risk dimensions X 1 ,..., X m P represents the state variables for each risk dimension. X 1 ,..., X m P( ) represents the joint probability of a combination of risk dimensions. X i ) represents the marginal probability of the i-th risk dimension state variable.

[0035] when T AA larger value indicates that the joint distribution of the corresponding risk dimension combination deviates more from the independent distribution, and the stronger the nonlinear coupling between risk factors. By comparing different risk combinations... T A The values ​​can be used to obtain the static coupling strength ranking and key static coupling modes.

[0036] Table 2 presents the ranking of static coupling strengths obtained in one embodiment. This result is used to illustrate that the method can distinguish the static coupling strength of different risky coupling modes, and is not intended to limit the scope of protection of this invention.

[0037] Table 2. Ranking of Static Coupling Strength Step S4: Set the baseline failure rate, correct the marginal probabilities of each risk factor based on the expert correction coefficient, and convert the corrected probabilities into the failure rate of the dynamic Bayesian network nodes.

[0038] Since the actual accident sample size of unmanned inland waterway vessels is limited, directly using the marginal probabilities of traditional ship accidents as the node probabilities of a dynamic Bayesian network may lead to prediction results that deviate from the unmanned vessel scenario. Therefore, this embodiment introduces an expert correction coefficient to calibrate the marginal probabilities.

[0039] Experts assessed the changes in risk factors for each risk factor in unmanned vessel scenarios compared to traditional vessel scenarios and provided revised scores. The [number]th... k The weight of renowned experts is recorded as follows: w k , and make it for the first i The revised score given for each risk factor is denoted as: r ik Then the first i Expert correction coefficient for each risk factor β i for: when β i A value greater than 1 indicates that the risk factor tends to increase in the context of unmanned vessels; when βi When the value is less than 1, it indicates that the risk factor tends to be reduced in the context of unmanned vessels.

[0040] No. i Correction probability of each risk factor for: Set system-level baseline failure rate , No. i Failure rate of individual risk factors for: The failure rate Used for calculating the time transition probability of risk factor nodes in subsequent dynamic Bayesian networks.

[0041] Table 3 shows the probability correction and failure rate calculation results for some risk factors in one embodiment.

[0042] Table 3 Probability Adjustments and Failure Rates for Some Risk Factors Step S5: Construct a dynamic Bayesian network model, set risk factor nodes, risk dimension nodes, and risk coupling nodes on continuous time slices, and determine the state transition probability between adjacent time slices.

[0043] like Figure 4 As shown, this embodiment constructs a multi-time-slice dynamic Bayesian network model. Each time slice includes risk factor nodes, risk dimension nodes, and risk coupling nodes. Risk factor nodes correspond to the secondary risk factors in step S1; risk dimension nodes correspond to the four primary risk dimensions C, P, E, and S; risk coupling nodes correspond to coupling modes such as CP, CE, CS, PE, PS, ES, CPE, CPS, CES, PES, and CPES; the overall risk node is used to characterize the overall failure risk state of unmanned inland waterway vessels.

[0044] The dynamic Bayesian network model employs the first-order Markov assumption, meaning that the node state in the current time slice is only related to the corresponding node state in the previous time slice and the state of the parent node within the current time slice. For the ... i The probability of a risk factor node transitioning from a normal state to a failure state is: If the first i If a risk factor node was already in a failed state in the previous time slice, then its failed state is set to the absorption state: in, =1 indicates the first i The risk factor node at the _ t One time slice is in an invalid state. =0 indicates a normal state, Δ t This indicates the interval between adjacent time slices.

[0045] It should be noted that the NK static coupling analysis and the DBN dynamic coupling analysis serve different functions in this invention: the NK model is mainly used to identify the static coupling strength and key coupling patterns of different risk dimension combinations; the DBN model is mainly used to infer the cumulative probability of the evolution of each risk factor and risk coupling pattern over time. The focus of this invention is to simultaneously output the results of static coupling analysis and dynamic coupling analysis.

[0046] Step S6: Perform dynamic coupling inference based on the dynamic Bayesian network model to output the cumulative failure probability under different time slices.

[0047] Based on the dynamic Bayesian network model constructed in step S5, forward reasoning is performed on the risk state within the preset time slice range to obtain the posterior probabilities of risk factor nodes, risk dimension nodes, risk coupling nodes, and overall risk nodes under different time slices.

[0048] No. i The risk factor node at the _ t Cumulative failure probability under a time slice for: in, Indicates the initial time slice. Represents a step function; when t < hour, =0, when t ≥ hour, =1.

[0049] like Figure 5 As shown, the output includes a static coupling strength ranking and a dynamic coupling evolution curve. The static coupling strength ranking is used to determine which combinations of risk dimensions have stronger nonlinear coupling effects; the dynamic coupling evolution curve is used to determine which risk coupling patterns grow faster over time and are more likely to become critical propagation paths.

[0050] Preferably, this embodiment can also perturb the failure rate of each risk factor node, calculate the change in the posterior probability of the overall risk node and the risk coupling node before and after the perturbation, and thus obtain the risk factor sensitivity ranking. This ranking can identify key risk factors that have a significant impact on the evolution of coupled risks of unmanned inland waterway vessels.

[0051] Example 2 Based on Example 1, this example provides a static-dynamic coupling risk analysis and evolution prediction system for inland waterway unmanned vessels, used to implement the static-dynamic coupling risk analysis and evolution prediction method for inland waterway unmanned vessels in Example 1. (See also...) Figure 6 The system includes: The risk indicator system construction module is used to construct a risk indicator system for unmanned inland waterway vessels. The risk indicator system includes four primary risk dimensions: algorithm, vessel physical entity, navigation environment, and shore-based control and management, as well as their respective secondary risk factors. The historical data mapping and probability calculation module is used to acquire historical accident data of inland waterway vessels, filter and functionally map the historical accident data to map the traditional causes of vessel accidents to the risk indicator system, generate a risk factor state matrix, and calculate the marginal probability of each risk factor and the joint probability of the combination of risk factors. The static coupling analysis module is used to encode the risk dimension state of each accident sample, calculate the two-factor, three-factor and four-factor risk coupling values ​​based on the NK model, and obtain the static coupling strength ranking and key static coupling patterns based on the risk coupling values. The failure rate determination module is used to set the baseline failure rate, introduce expert correction coefficients to correct the marginal probabilities of risk factors, and convert the corrected probabilities into the failure rate of dynamic Bayesian network nodes. The Dynamic Bayesian Network Building Module is used to build a multi-time-slice dynamic Bayesian network model. It sets risk factor nodes, risk dimension nodes, and risk coupling nodes in each time slice and determines the state transition probability between adjacent time slices based on the failure rate. The dynamic coupling inference module is used for forward inference based on the dynamic Bayesian network model, outputting the cumulative failure probability of each node under different time slices, and generating dynamic coupling evolution curves. The sensitivity analysis module is used to perturb the failure rate of each risk factor node, calculate the change in the posterior probability of the overall risk node and risk coupling node before and after the perturbation, and obtain the sensitivity ranking of risk factors based on the change.

[0052] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention uses the NK model to perform static coupling analysis on the combination of risk dimensions, which can quantify the coupling strength of two-factor, three-factor and four-factor risks, and avoid simply attributing complex accidents to a single risk factor.

[0053] 2. This invention uses dynamic Bayesian networks to infer the state transition and cumulative failure process of risk factors in continuous time slices, and can characterize the dynamic evolution of risk coupling relationships with running time.

[0054] 3. This invention combines historical accident data, functional mapping, and expert correction, enabling the generation of node failure rates that can be used for dynamic reasoning even when there are insufficient actual operational samples of unmanned inland waterway vessels.

[0055] 4. This invention can simultaneously output static coupling strength ranking and dynamic coupling evolution path, providing a basis for risk identification, risk warning, maintenance priority determination and ship-shore collaborative management of unmanned inland waterway vessels.

[0056] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for static and dynamic coupling risk analysis and evolution prediction of unmanned inland waterway vessels, characterized in that, Includes the following steps: Construct a risk indicator system for unmanned inland waterway vessels, dividing the risk indicators into primary risk dimensions and secondary risk factors; The system acquires historical accident data of inland waterway vessels, filters and maps the historical accident data, maps the traditional causes of vessel accidents to the risk indicator system, generates a risk factor state matrix, and calculates the marginal probability of each risk factor and the joint probability of the combination of risk factors. Based on the NK model, two-factor, three-factor, and four-factor risk coupling values ​​are calculated, and static coupling strength ranking and key static coupling patterns are obtained according to the risk coupling values. A baseline failure rate is set, and the marginal probabilities of the risk factors are corrected based on a preset expert correction coefficient. The corrected probabilities are then converted into the failure rate of the dynamic Bayesian network nodes. A dynamic Bayesian network model is constructed, and risk factor nodes, risk dimension nodes, and risk coupling nodes are set on continuous time slices. The state transition probability between adjacent time slices is determined based on the failure rate. Based on the dynamic Bayesian network model, dynamic coupling inference is performed to output the cumulative failure probability under different time slices and generate dynamic coupling evolution curves.

2. The method for static and dynamic coupling risk analysis and evolution prediction of unmanned inland waterway vessels according to claim 1, characterized in that, In the dynamic Bayesian network model, the risk-coupled nodes include CP, CE, CS, PE, PS, ES, CPE, CPS, CES, PES, and CPES.

3. The method for static and dynamic coupling risk analysis and evolution prediction of unmanned inland waterway vessels according to claim 1, characterized in that, The process of correcting the marginal probabilities of the risk factors based on preset expert correction coefficients includes: The first The weight of renowned experts is recorded as follows: , and make it for the first The adjusted score for each risk factor is denoted as: Then the first Expert correction coefficients for each risk factor ; The corrected probability is ; Failure rate of the dynamic Bayesian network nodes ,in This is the system-level baseline failure rate.

4. The method for static and dynamic coupling risk analysis and evolution prediction of unmanned inland waterway vessels according to claim 3, characterized in that, The state transition probability between adjacent time slices is: For the Risk factor nodes Transition from normal state To failure state probability If the slice was already in an invalid state at a previous time, then ,in The dynamic Bayesian network model employs the first-order Markov assumption, with adjacent time slice intervals.

5. The method for static and dynamic coupling risk analysis and evolution prediction of unmanned inland waterway vessels according to claim 1, characterized in that, The risk coupling value The calculation formula is: in, Indicates by A combination of risk dimensions Representing the state variables of each risk dimension, The joint probability of the combination of risk dimensions. For the first Marginal probabilities of each risk dimension.

6. The method for static and dynamic coupling risk analysis and evolution prediction of unmanned inland waterway vessels according to claim 5, characterized in that, The process of obtaining the static coupling strength ranking and key static coupling patterns based on the risk coupling value includes the following steps: Calculate the risk coupling values ​​for two-factor, three-factor, and four-factor scenarios, compare the magnitudes of each risk coupling value to obtain a ranking of static coupling strength, and identify the combination with the larger risk coupling value as the key static coupling mode.

7. The method for static and dynamic coupling risk analysis and evolution prediction of unmanned inland waterway vessels according to claim 1, characterized in that, The risk factor state matrix is ​​composed of the risk factor state vectors of each accident sample. For the first... j The accident and the first i One risk factor; if the risk factor exists, then the state value... x ij = 1, otherwise x ij = 0; The marginal probability is ,in The total number of accident samples. For including the first i Number of accident samples for each risk factor; The combination of risk factors A joint probability ,in To simultaneously include combinations A The number of accident samples for each risk factor in China.

8. The method for static and dynamic coupling risk analysis and evolution prediction of unmanned inland waterway vessels according to claim 1, characterized in that, The function mapping includes: In traditional ship accidents, improper lookout is mapped to a failure of perception logic; misjudgment or improper collision avoidance decision is mapped to a failure of decision planning; improper maneuvering is mapped to a failure of motion control logic; equipment failure is mapped to risk factors related to the ship's physical entity; meteorological, hydrological, and waterway constraints are mapped to risk factors related to the navigation environment; and insufficient management, supervision, communication, and remote support are mapped to risk factors related to shore-based control and management.

9. The method for static and dynamic coupling risk analysis and evolution prediction of unmanned inland waterway vessels according to claim 1, characterized in that, The primary risk dimensions include algorithms, ship physical entities, navigation environment, and shore-based control and management; among which... The algorithm dimension includes failures in perception logic, decision planning, motion control logic, and defects in algorithm training data; The physical entity dimension of a ship includes failures of power and actuators, failures of sensing and communication hardware, failures of onboard computing platforms, and failures of automatic loading and stability. The navigation environment dimension includes severe weather interference, complex hydrological interference, and restricted navigation environment; The shore-based control and management dimensions include remote monitoring takeover failures, cybersecurity and management vulnerabilities, and insufficient remote maintenance support.

10. A static-dynamic coupling risk analysis and evolution prediction system for unmanned inland waterway vessels, characterized in that, The system is used to implement the static and dynamic coupling risk analysis and evolution prediction method for inland waterway unmanned vessels as described in any one of claims 1-9, and includes: The risk indicator system construction module is used to construct a risk indicator system for unmanned inland waterway vessels. The risk indicator system includes four primary risk dimensions: algorithm, vessel physical entity, navigation environment, and shore-based control and management, as well as their respective secondary risk factors. The historical data mapping and probability calculation module is used to acquire historical accident data of inland waterway vessels, filter and functionally map the historical accident data to map the traditional causes of vessel accidents to the risk indicator system, generate a risk factor state matrix, and calculate the marginal probability of each risk factor and the joint probability of the combination of risk factors. The static coupling analysis module is used to encode the risk dimension state of each accident sample, calculate the two-factor, three-factor and four-factor risk coupling values ​​based on the NK model, and obtain the static coupling strength ranking and key static coupling patterns based on the risk coupling values. The failure rate determination module is used to set a baseline failure rate, introduce an expert correction coefficient to correct the marginal probability of the risk factor, and convert the corrected probability into the failure rate of the dynamic Bayesian network node. The dynamic Bayesian network construction module is used to construct a multi-time-slice dynamic Bayesian network model. In each time slice, risk factor nodes, risk dimension nodes, and risk coupling nodes are set, and the state transition probability between adjacent time slices is determined based on the failure rate. The dynamic coupling inference module is used to perform forward inference based on the dynamic Bayesian network model, output the cumulative failure probability of each node under different time slices, and generate dynamic coupling evolution curves. The sensitivity analysis module is used to perturb the failure rate of each risk factor node, calculate the change in the posterior probability of the overall risk node and risk coupling node before and after the perturbation, and obtain the sensitivity ranking of risk factors based on the change.

Images