Shipping man-machine coupling failure risk assessment method and system based on graph theory

Abstract

The invention provides a graph theory-based shipping man-machine coupling failure risk assessment method and system, and relates to the technical field of shipping risk assessment, and the method comprises the steps: extracting structured data with a timestamp, generating an effective data set, screening the effective data set, and generating a key risk factor set; constructing a weighted undirected network through a graph theory based on the key risk factor set, performing sliding segmentation on the effective data set through setting time windows with fixed lengths and a preset step length, recalculating a standardized mutual information value between nodes in each time window, updating an edge weight, and generating a time-varying weighted undirected network topology sequence; performing global efficiency analysis on the time-varying weighted undirected network topology sequence, calculating a network global efficiency decline rate after each node is removed, and sorting the key risk factors according to the decline rate to determine the priority of the key risk factors; and outputting corresponding accident prevention suggestions and sailor training schemes based on the key risk factor priorities.

Description

Technical Field

[0001] This invention relates to the field of shipping risk assessment technology, and in particular to a graph theory-based method and system for assessing shipping human-machine coupling failure risks. Background Technology

[0002] With the rapid development of autonomous ship technology, the level of automation and intelligence of ships is constantly improving, further increasing the complexity of human-machine interaction. The operation of autonomous ships relies on highly complex human-machine interaction systems, in which the relationship between human factors and machine malfunctions is becoming increasingly intricate.

[0003] However, most existing maritime risk assessment models rely on static validation methods and lack mechanisms for evaluating their adaptability and reliability in dynamic environments. Maritime accidents are often accompanied by unforeseen circumstances and environmental changes, and traditional models exhibit low reliability in the face of dynamic changes. The difficulty in capturing the temporal evolution of relationships between risk factors limits the accurate assessment of dynamic risk situations. Summary of the Invention

[0004] The purpose of this invention is to provide a graph theory-based method and system for assessing the failure risk of human-machine coupling in shipping, in order to solve the problem mentioned in the background art that most existing maritime risk assessment models rely on static verification methods, which make it difficult to capture the temporal evolution characteristics of the correlation between risk factors and limit the accurate assessment of dynamic risk situations.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a graph theory-based method for assessing the risk of human-machine coupling failure in shipping, comprising the following steps: extracting structured data including timestamps from a shipping accident database, performing data cleaning and standardization to generate an effective dataset, wherein the structured data includes human factor data, machine failure data, environmental factor data, and management factor data; filtering the effective dataset and generating a set of key risk factors; constructing a weighted undirected network based on the set of key risk factors using graph theory, performing sliding partitioning of the effective dataset by setting a fixed-length time window and a preset step size, recalculating the standardized mutual information values ​​between nodes and updating edge weights within each time window, thereby generating a time-varying weighted undirected network topology sequence; performing global efficiency analysis on the time-varying weighted undirected network topology sequence, calculating the global efficiency decline rate of the network after each node is removed, and ranking the key risk factors according to the decline rate to determine the priority of the key risk factors; and outputting corresponding accident prevention suggestions and crew training programs based on the priority of the key risk factors.

[0006] Optionally, the step of filtering the effective dataset and generating a set of key risk factors specifically includes: filtering risk factors related to machine failure in the effective dataset using a chi-square test to remove independent variables; performing a multicollinearity test on the filtered risk factors using LASSO regression to remove redundant variables; and using a causal random forest algorithm to mine the nonlinear causal association between the set of key risk factors and the machine failure data to obtain the set of key risk factors.

[0007] Optionally, the step of screening risk factors related to machine failure in the effective dataset using the chi-square test specifically includes: performing preliminary screening of the effective dataset using the chi-square test; establishing a null hypothesis that the risk factors and machine failure are independent, and an alternative hypothesis that the risk factors and machine failure are not independent; constructing a contingency table of the risk factors and the machine failure, calculating the chi-square value based on the observed frequency and theoretical frequency in the contingency table, and comparing it with a preset critical value; rejecting the null hypothesis when the chi-square value is greater than the critical value, and preliminarily screening the risk factor as a candidate key factor for subsequent analysis.

[0008] Optionally, the step of using LASSO regression to test for multicollinearity of the screened risk factors specifically includes: setting a regularization parameter, constraining the regression coefficient by minimizing the objective function, and removing a candidate key factor when its regression coefficient shrinks to zero, thereby eliminating multicollinear variables.

[0009] Optionally, the step of using the causal random forest algorithm to mine the nonlinear causal relationship between the set of key risk factors and the machine failure data specifically includes: constructing multiple decision trees to form a forest structure, with each tree trained based on a random feature subset; setting machine failure as the dependent variable, calculating the importance score of the variable through heterogeneity weighting, and screening key risk factors whose scores exceed a preset first threshold and whose causal effect values ​​exceed a preset second threshold.

[0010] Optionally, the step of constructing a weighted undirected network based on the set of key risk factors using graph theory specifically includes: using each factor in the set of key risk factors as a node, calculating the correlation strength between nodes based on standardized mutual information as edge weights, and constructing a weighted undirected network.

[0011] Optionally, the step of determining the priority of key risk factors specifically includes: calculating the shortest path distance between all node pairs in the time-varying weighted undirected network topology sequence, quantifying the global efficiency among risk factors using the global efficiency formula, and using this as a benchmark value; sequentially removing individual nodes and all their associated edges to form sub-networks, and recalculating the global efficiency of each sub-network; calculating the global efficiency decline rate corresponding to each node based on the difference between the benchmark value and the global efficiency of each sub-network; and sorting the key risk factors in descending order according to the global efficiency decline rate to determine the priority of key risk factors.

[0012] On the other hand, the present invention also provides a graph theory-based shipping human-machine coupling failure risk assessment system, comprising: an extraction module for extracting structured data including timestamps from a shipping accident database, performing data cleaning and standardization to generate an effective dataset, wherein the structured data includes risk factor data and machine failure data; a key factor generation module for filtering the effective dataset and generating a set of key risk factors; a time-varying network generation module for constructing a weighted undirected network based on the set of key risk factors using graph theory, performing sliding partitioning of the effective dataset by setting a fixed-length time window and a preset step size, recalculating the standardized mutual information values ​​between nodes and updating edge weights within each time window, thereby generating a time-varying weighted undirected network topology sequence; a priority module for performing global efficiency analysis on the time-varying weighted undirected network topology sequence, calculating the global efficiency decline rate of the network after each node is removed, and sorting the key risk factors according to the decline rate to determine the priority of the key risk factors; and an output module for outputting corresponding accident prevention suggestions and crew training programs based on the priority of the key risk factors.

[0013] On the other hand, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described graph theory-based shipping human-machine coupling failure risk assessment method.

[0014] On the other hand, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described graph theory-based shipping human-machine coupling failure risk assessment method.

[0015] Compared with the prior art, the beneficial effects of the present invention are: This application extracts and cleans time-stamped structured data from shipping accident data, filters the effective dataset, and generates a set of key risk factors. This effectively removes interference from irrelevant and redundant variables, focuses on the core elements that truly affect system risk, and improves the efficiency and accuracy of the analysis.

[0016] This application constructs a time-varying weighted undirected network based on graph theory and generates a topology sequence, realizing the quantification of the dynamic correlation between various risk factors in the shipping human-machine coupling system. It can identify the temporal variation patterns of seasonal risk patterns, the impact of sudden events, and long-term trend changes, providing a refined network foundation for dynamic risk assessment.

[0017] This application performs a global efficiency analysis and calculates the efficiency decline rate after node removal. Based on this indicator, risk factors are prioritized, enabling a scientific quantification of the impact of each factor on the overall system stability and identifying the most critical weaknesses. Finally, based on these priorities, incident prevention recommendations and training programs are output, significantly improving the relevance and effectiveness of safety measures. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the method steps of the present invention.

[0019] Figure 2 This is a flowchart of the method of the present invention.

[0020] Figure 3 This is a schematic diagram of the system structure of the present invention.

[0021] In the diagram: 10 - Extraction module, 20 - Key factor generation module, 30 - Time-varying network generation module, 40 - Priority module, 50 - Output module. Detailed Implementation

[0022] The present invention will now be clearly and completely described in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0023] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be used interchangeably where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0024] Those skilled in the art will understand that, unless explicitly stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in the specification of this application means the presence of features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0025] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0026] It should be understood that the sequence number and size of each step in this embodiment do not imply the order of execution. The execution order of each process is determined by its function and internal logic, and should not constitute any limitation on the implementation process of this application embodiment.

[0027] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0028] Please refer to Figures 1-2This invention discloses a graph theory-based method for assessing the risk of human-machine coupling failure in shipping. The method includes the following steps: extracting structured data with timestamps from a shipping accident database, performing data cleaning and standardization to generate a valid dataset, wherein the structured data includes human factor data, machine failure data, environmental factor data, and management factor data; filtering the valid dataset to generate a set of key risk factors; constructing a weighted undirected network based on the key risk factor set using graph theory, performing sliding partitioning of the valid dataset by setting a fixed-length time window and a preset step size, recalculating the standardized mutual information values ​​between nodes and updating edge weights within each time window, thereby generating a time-varying weighted undirected network topology sequence; performing global efficiency analysis on the time-varying weighted undirected network topology sequence, calculating the global efficiency decline rate after each node is removed, and ranking the key risk factors according to the decline rate to determine their priority; and outputting corresponding accident prevention suggestions and crew training programs based on the priority of the key risk factors.

[0029] Specifically, accident reports are collected from the global maritime accident database. Preliminary screening of data is conducted by selecting the time span of collection, accident location, vessel type, or language of the accident report to ensure the authority and completeness of the data source. Accident reports are then manually reviewed to filter out data with missing information, unclear causes, or overly simplistic reports. Missing values ​​are handled using deletion or interpolation methods to ensure data quality meets analytical requirements. Based on the International Maritime Organization classification standards and key elements in accident reports, risk factors are standardized and classified to establish a multi-dimensional risk factor system encompassing human factors, machinery failure, environmental factors, and management factors.

[0030] Based on global shipping accident analysis standards and key elements of historical accident reports, human factors in structured data are fully categorized, including: lack of crew knowledge, improper use of equipment, lack of equipment inspection, equipment shortage, observation errors, psychological stress, confusion, negligence in cargo inspection, and insufficient crew qualifications. Each category of factors is directly related to the causal mechanism of ship machinery failure, providing a complete risk factor framework for subsequent quantitative analysis.

[0031] The human factors mentioned include: lack of knowledge among crew members, improper use of equipment, lack of equipment inspection, observation errors, psychological stress, confusion, negligence in cargo inspection, insufficient crew qualifications, lack of protection, ineffective supervision, communication failures, improper operation, and lack of equipment.

[0032] The chi-square test was used to initially screen risk factors that were significantly associated with machine failure, establish the independence hypothesis, and remove irrelevant variables through significance judgment. LASSO regression was used to test the multicollinearity of the initial screening factors, and redundant variables were removed through regularization constraints. The causal random forest algorithm was used to mine the nonlinear causal relationship between risk factors and machine failure, focusing on the core factors that have a prominent impact on human-machine coupling risk.

[0033] Based on graph theory, a time-varying weighted undirected network is constructed and a topology sequence is generated. This is the first time that the dynamic relationships between various risk factors in a shipping human-machine coupled system have been quantified. It can identify the temporal variations of seasonal risk patterns, the impact of sudden events, and long-term trend changes, providing a refined network foundation for dynamic risk assessment. Global efficiency analysis is performed, and the efficiency decline rate after node removal is calculated. Based on this indicator, risk factors are prioritized, enabling a scientific quantification of the impact of each factor on the overall system stability and identifying the most critical weak points. Finally, based on these priorities, accident prevention recommendations and training programs are output, providing intuitive and quantitative data support for risk management decisions and resource allocation, significantly improving the pertinence and effectiveness of safety measures.

[0034] In some embodiments, the step of filtering the effective dataset and generating a set of key risk factors specifically includes: filtering risk factors related to machine failure in the effective dataset using a chi-square test to remove independent variables; performing a multicollinearity test on the filtered risk factors using LASSO regression to remove redundant variables; and using a causal random forest algorithm to mine the nonlinear causal association between the set of key risk factors and the machine failure data to obtain the set of key risk factors.

[0035] Specifically, when performing chi-square tests on each human factor field, the null hypothesis "human factors and machine failures are independent" and the alternative hypothesis "human factors and machine failures are not independent" are established; theoretical frequencies are calculated based on the contingency table, and the chi-square value is calculated using the formula, then compared with the significance level. The corresponding critical value is compared, and when the chi-square value is greater than the critical value, the null hypothesis is rejected and the factor is determined to be a key human factor.

[0036] The application of LASSO regression in human-machine coupling failure analysis in shipping systems involves: setting screening thresholds for multi-dimensional human factors such as crew operational behavior, equipment maintenance status, and environmental awareness; identifying and eliminating redundant factors with low contribution to machine failures by constraining regression coefficients; and focusing on core human-machine interaction failure elements that significantly impact shipping safety. These selected key human-machine coupling failure influencing factors will serve as core elements of the failure correlation graph, laying the foundation for constructing a weighted correlation network reflecting the coupling relationship between personnel operations and equipment responses in shipping operations.

[0037] This study utilizes a causal random forest algorithm to uncover nonlinear causal relationships between risk factors and machine failures. A forest structure is formed by constructing multiple decision trees, each trained based on a random subset of features, thus avoiding overfitting.

[0038] This application employs a multi-level, progressive screening process using chi-square test, LASSO regression, and causal random forest algorithm to sequentially address the issues of variable independence screening, multicollinearity elimination, and nonlinear causal association mining. This combined screening method significantly improves the purity and representativeness of the key risk factor set, aiming to solve two major deficiencies in existing shipping risk assessments: first, the inability to objectively model the interaction process between personnel and machines in shipping systems and their complex dependencies, making it difficult to characterize the dynamic correlation between human factors and machine failures in shipping accidents; and second, the difficulty of traditional risk analysis in quantifying the risk assessment of human-machine coupling failures in shipping systems from a global perspective, with redundant variables in the model affecting the accuracy of key human factor identification and the robustness of the model, making it difficult to derive strategies to support the prevention and control of human-machine coupling failure risks in actual shipping systems.

[0039] In some embodiments, the step of screening risk factors related to machine failure in the effective dataset using the chi-square test specifically includes: performing preliminary screening of the effective dataset using the chi-square test; establishing a null hypothesis that the risk factors and machine failure are independent of each other, and an alternative hypothesis that the risk factors and machine failure are not independent of each other; constructing a contingency table of the risk factors and the machine failure, calculating the chi-square value based on the observed frequency and theoretical frequency in the contingency table, and comparing it with the critical value corresponding to the significance level α=0.05; rejecting the null hypothesis when the chi-square value is greater than the critical value, and preliminarily screening the risk factor as a candidate key factor for subsequent analysis.

[0040] Specifically, the chi-square test method was used to screen risk factors significantly associated with machine failure: First, a statistical hypothesis system was established, where the null hypothesis H0 was defined as the target risk factor and machine failure being independent of each other, and the alternative hypothesis H1 was defined as the two being significantly associated. A significance level was set during the testing process. Determining degrees of freedom based on the characteristics of binary variables The corresponding chi-square distribution critical value .

[0041] For each risk factor, construct a 2×2 contingency table and calculate the theoretical frequency. Chi-square statistic: ; ; In the formula: For the frequency of observation, and Do not use row or column marginal frequencies. This represents the total sample size. When calculated... If the value is greater than the critical value of 3.841, reject H0 and retain the risk factor.

[0042] This application provides an objective and quantifiable statistical test standard for the initial screening of risk factors related to machine failure by establishing clear statistical hypotheses and using contingency tables to calculate chi-square values ​​for significance assessment. This method effectively avoids biases that may arise from subjective experience, ensuring the scientific rigor and precision of the initial screening process, and providing a high-quality set of candidate variables for subsequent, more refined screening stages.

[0043] In some embodiments, the step of using LASSO regression to test for multicollinearity of the screened risk factors specifically includes: setting a regularization parameter, constraining the regression coefficients by minimizing the objective function, and eliminating a candidate key factor when its regression coefficient shrinks to zero, thereby eliminating multicollinear variables.

[0044] Specifically, LASSO regression was used to test for multicollinearity of the factors screened by the chi-square test, and redundant variables were eliminated by introducing L1 regularization constraints. The objective function of LASSO regression is: ; In the formula: y is the machine fault label vector, X is the risk factor matrix, β is the regression coefficient vector, and λ is the regularization parameter. The L1 norm is used. The optimal regularization parameter λ is determined by cross-validation. When the regression coefficient of a variable shrinks to zero under the regularization constraint, the variable is removed from the feature set, thereby effectively eliminating the interference of multicollinear variables.

[0045] This application introduces LASSO regression and sets regularization parameters to constrain the regression coefficients, automatically compressing and eliminating the coefficients of variables exhibiting multicollinearity and redundant contributions to the model. This technique effectively solves the common problem of high-dimensional variable correlation interference in maritime risk data, simplifies the model structure, enhances the model's interpretability and generalization ability, and avoids distortion of evaluation results caused by multicollinearity of dependent variables.

[0046] In some embodiments, the step of using the causal random forest algorithm to mine the nonlinear causal association between the set of key risk factors and the machine failure data specifically includes: constructing multiple decision trees to form a forest structure, with each tree trained based on a random feature subset; setting machine failure as the dependent variable, calculating the variable importance score through heterogeneity weighting, and screening key risk factors whose scores exceed a preset first threshold and whose causal effect values ​​exceed a preset second threshold.

[0047] Specifically, the causal random forest algorithm is used to uncover nonlinear causal relationships between risk factors and machine failures. A forest structure is formed by constructing multiple decision trees, each trained based on a random subset of features to avoid overfitting. Causal effect estimation is achieved through conditional averaging of effects. ; In the formula: Y1 and Y0 represent the outcome variables of the treatment group and the control group, respectively, and X is a covariate. The importance score and causal effect value of each risk factor are calculated by heterogeneity weighting. A reasonable screening threshold θ and a minimum causal effect value β are set to screen risk factors with scores exceeding the threshold and significant causal effect values. Finally, the set of key risk factors is output as the node basis for subsequent network modeling.

[0048] This application constructs multiple decision trees using a causal random forest algorithm and calculates variable importance scores and causal effect values, enabling in-depth exploration of the complex nonlinear causal relationships between risk factors and machine failures. This method overcomes the limitations of traditional linear models in handling complex human-machine system interactions, identifying key risk factors with strong causal drivers, thus making the attribution analysis of human-machine coupling failure risks more in-depth and accurate.

[0049] In some embodiments, the step of constructing a weighted undirected network based on the set of key risk factors using graph theory specifically includes: using each factor in the set of key risk factors as a node, calculating the correlation strength between nodes based on standardized mutual information as edge weights, and constructing a weighted undirected network.

[0050] Specifically, graph theory networks It consists of a set of nodes V and a set of edges E, where nodes represent key risk factors and edges represent the relationships between factors. By introducing time-varying characteristics, the network can dynamically reflect the evolution of the relationships between risk factors over time, providing theoretical support for dynamic risk assessment of human-machine coupled systems.

[0051] The key risk factors identified are strictly defined as network nodes, with each node... This system characterizes the state features of a single risk factor in a maritime accident. Node attribute design includes three dimensions: factor name identifier, state type description, and timestamp. The state type uses a binary encoding method, where the presence of the risk factor in a specific accident is 1, and its absence is 0. The timestamp provides a temporal basis for constructing the time-varying network, ensuring that the network topology accurately reflects the dynamic correlation patterns of risk factors across different time periods.

[0052] Standardized mutual information is used as the edge weight calculation method to accurately quantify the strength of nonlinear associations between nodes. Standardized mutual information can effectively handle nonlinear dependencies between variables, avoiding the limitations of traditional linear correlation coefficients. The calculation formula is as follows: ; ; ; In the formula: Mutual information value; and Variables , Information entropy.

[0053] Through standardization, the range of standardized mutual information is restricted to the interval [0,1], where 0 represents complete independence and 1 represents complete correlation, ensuring that the edge weights have good comparability and mathematical properties.

[0054] This application successfully transforms abstract risk relationships into a concrete network topology by explicitly defining each key risk factor as a network node and using standardized mutual information to calculate the correlation strength between nodes as edge weights. This transformation makes the interactions between elements within a complex system visible and measurable, providing a direct model foundation for using graph theory methods to conduct systemic risk analysis.

[0055] In some embodiments, to capture the temporal evolution characteristics of the correlation between risk factors, this invention designs a sliding time window mechanism to construct a time-varying network topology. A fixed-length time window T and a sliding step size Δt are set, and the maritime accident data is segmented according to the time series. Within each time window... Within this process, the standardized mutual information values ​​between all node pairs are recalculated, and the network edge weight matrix is ​​updated. : ; A time-varying network sequence {G(t1), G(t2), ..., G(t3)} is generated using a sliding window mechanism. Each network snapshot G( This reflects the correlation patterns of risk factors within a specific time period. This time-varying topology can identify seasonal risk patterns, the impact of sudden events, and long-term trend changes, providing a refined network foundation for dynamic risk assessment.

[0056] The constructed time-varying weighted undirected network possesses the following important properties: First, the network's undirectedness reflects the symmetric characteristics of the mutual influence among risk factors; second, the non-negativity of edge weights ensures the physical meaning of the association strength; third, the time-varying nature enables the network to adapt to the dynamic changes of complex maritime environments. Analysis of topological indices such as network density, clustering coefficient, and average path length verifies the rationality and effectiveness of the network model, laying a mathematical foundation for subsequent global efficiency analysis.

[0057] In some embodiments, the step of determining the priority of key risk factors specifically includes: calculating the shortest path distance between all node pairs in the time-varying weighted undirected network topology sequence, quantifying the global efficiency among risk factors using the global efficiency formula, and using this as a benchmark value; sequentially removing individual nodes and all their associated edges to form sub-networks, and recalculating the global efficiency of each sub-network; calculating the global efficiency decline rate corresponding to each node based on the difference between the benchmark value and the global efficiency of each sub-network; and sorting the key risk factors in descending order according to the global efficiency decline rate to determine the priority of key risk factors.

[0058] Specifically, global network efficiency is an important indicator for measuring the information transmission capacity of complex networks. It quantifies the overall connectivity and synergistic effect between nodes in the network based on the shortest path theory. In the risk network of a shipping human-machine coupled system, global efficiency reflects the overall strength and propagation efficiency of the interactions between risk factors. High efficiency values ​​indicate that risk factors in the network are closely related, information transmission paths are short, and propagation speed is fast, predicting the rapid spread and amplification effect of system risks.

[0059] The global efficiency of the network is calculated using the shortest path algorithm in graph theory. First, the Floyd-Warshall algorithm is used to determine the shortest path distance between any two nodes in the network. In a weighted network, the path distance is defined as the sum of the reciprocals of the weights of all edges along the path, ensuring that high correlation strength corresponds to short distances. The formula for calculating the global network efficiency is: ; In the formula: N is the total number of network nodes. Let be the shortest path distance between nodes i and j. Defined when there is no connected path between the two nodes. Corresponding The global efficiency range is [0,1], where 0 indicates that the network is completely disconnected, and 1 indicates that the network is a complete graph and all edge weights are at their maximum values.

[0060] To quantify the contribution of each risk factor to the overall system risk, this invention designs a node elimination experiment to evaluate the importance of individual nodes. The experimental process includes the following steps: First, the global efficiency of the original complete network is calculated. Used as a baseline; then remove individual nodes from the network one by one. Combine the node and all its associated edges to form a subnetwork of the missing node i; recalculate the global efficiency of the subnetwork. ;Calculate the efficiency degradation rate after node i is removed: ; Among them, the efficiency decline rate It directly reflects the degree of contribution of node i to the overall network connectivity. The greater the decline rate, the more important the core position of the risk factor in the human-machine coupling system.

[0061] Based on the results of the node elimination experiment, according to the efficiency reduction rate The numerical values ​​of all risk factors are sorted in descending order to generate a quantified risk priority list. This ranking identifies the most critical core factors impacting system risk, providing a scientific basis for the optimal allocation of limited resources. Furthermore, by measuring the differences in efficiency decline rates, the relative importance of different risk factors can be quantified, avoiding the subjectivity and ambiguity of traditional qualitative assessments.

[0062] By combining time-varying network topology, dynamic tracking and analysis of risk priorities within different time windows are performed. By comparing changes in priority ranking across different periods, the temporal evolution and periodic characteristics of risk factor importance are identified. A Priority Change Index (PCI) is established. ; In the formula: Let i be the priority ranking of risk factors at time t. Let N be the priority ranking of risk factor i at the previous moment, and N be the total number of factors. This index can quantify the temporal stability of priority, providing a time-series basis for the formulation of dynamic risk management strategies.

[0063] This application quantifies the negative impact of a single risk factor failure on the overall efficiency of the network through systematic node elimination experiments. This result directly reflects the criticality and influence of a specific factor in the entire human-machine coupled risk network, thus providing an objective and quantitative basis for accurate risk factor classification and importance ranking. This allows limited security resources to be prioritized for controlling the risk points that pose the greatest threat to system stability.

[0064] In some embodiments, targeted decision support solutions are generated based on the quantification results of risk priority. For the key risk factors with the highest efficiency decline rate, operational standard enhancement measures and technical improvement plans are formulated; a tiered risk defense mechanism is established, prioritizing the allocation of limited safety resources to the control of high-priority risk factors; and, in conjunction with the trend of ship automation development, a crew training module for human-machine collaborative scenarios is designed, focusing on improving crew members' ability to identify and respond to key risk factors. Simultaneously, a dynamic early warning mechanism is established based on the results of time-varying priority analysis. When the importance of a specific risk factor fluctuates abnormally, a timely warning signal is issued and corresponding emergency response procedures are initiated, providing comprehensive technical support for intelligent safety management in the modern shipping industry.

[0065] This application utilizes risk factor priorities derived from time-varying weighted undirected network analysis to further optimize the design of human-machine collaborative systems. By analyzing the distribution and temporal evolution of the influence of key nodes revealed by global efficiency analysis, the targeted modules of the crew training framework are reconstructed. Combined with the functional topology of the autonomous vessel's operating interface, high-impact risk factors are mapped to key areas of human-machine interaction, establishing an early warning mechanism for risk factor correlation patterns. This enables collaborative decision support with risk visualization, providing decision support for automated vessel accident prevention.

[0066] Please refer to Figure 3 On the other hand, the present invention also provides a graph theory-based shipping human-machine coupling failure risk assessment system, comprising: an extraction module for extracting structured data including timestamps from a shipping accident database, performing data cleaning and standardization to generate an effective dataset, wherein the structured data includes human factor data, machine failure data, environmental factor data, and management factor data; a key factor generation module for filtering the effective dataset and generating a set of key risk factors; a time-varying network generation module for constructing a weighted undirected network based on the set of key risk factors using graph theory, performing sliding partitioning of the effective dataset by setting a fixed-length time window and a preset step size, recalculating the standardized mutual information values ​​between nodes and updating edge weights within each time window, thereby generating a time-varying weighted undirected network topology sequence; a priority module for performing global efficiency analysis on the time-varying weighted undirected network topology sequence, calculating the global efficiency decline rate of the network after each node is removed, and sorting the key risk factors according to the decline rate to determine the priority of the key risk factors; and an output module for outputting corresponding accident prevention suggestions and crew training programs based on the priority of the key risk factors.

[0067] On the other hand, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described graph theory-based shipping human-machine coupling failure risk assessment method.

[0068] On the other hand, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described graph theory-based shipping human-machine coupling failure risk assessment method.

[0069] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0070] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, database, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0071] The above are merely embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention's specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A graph theory-based method for assessing the failure risk of human-machine coupling in shipping, characterized by the following steps: include: Structured data, including timestamps, is extracted from the shipping accident database and then cleaned and standardized to generate a valid dataset. The structured data includes data on human factors, machine failures, environmental factors, and management factors. The valid dataset is filtered, and a set of key risk factors is generated; Based on the set of key risk factors, a weighted undirected network is constructed using graph theory. The effective dataset is then partitioned by setting a fixed-length time window and a preset step size. Within each time window, the standardized mutual information values ​​between nodes are recalculated and the edge weights are updated, thereby generating a time-varying weighted undirected network topology sequence. A global efficiency analysis is performed on the time-varying weighted undirected network topology sequence. The global efficiency reduction rate of the network after each node is removed is calculated, and the key risk factors are ranked according to the reduction rate to determine the priority of the key risk factors. Based on the priority of the key risk factors, corresponding accident prevention recommendations and crew training programs are provided.

2. The graph theory-based method for assessing the failure risk of human-machine coupling in shipping, as described in claim 1, is characterized in that... The steps of filtering the valid dataset and generating a set of key risk factors specifically include: The chi-square test was used to screen the effective dataset for risk factors related to machine failure in order to eliminate independent variables; LASSO regression was used to perform multicollinearity tests on the screened risk factors to eliminate redundant variables; The nonlinear causal relationship between the set of key risk factors and the machine failure data is obtained by using the causal random forest algorithm.

3. The graph theory-based method for assessing the failure risk of human-machine coupling in shipping, as described in claim 2, is characterized in that... The step of screening risk factors related to machine failure in the valid dataset using the chi-square test specifically includes: The effective dataset was initially screened using a chi-square test; Establish the null hypothesis that risk factors and machine failures are independent of each other, and the alternative hypothesis that risk factors and machine failures are not independent of each other; Construct a contingency table of the risk factors and the machine failures, calculate the chi-square value based on the observed frequency and theoretical frequency in the contingency table, and compare it with a preset critical value; When the chi-square value is greater than the critical value, the null hypothesis is rejected, and the risk factor is initially screened as a candidate key factor for subsequent analysis.

4. The graph theory-based method for assessing the failure risk of human-machine coupling in shipping, as described in claim 2, is characterized in that... The steps for performing multicollinearity testing on the screened risk factors using LASSO regression specifically include: By setting regularization parameters and constraining the regression coefficients by minimizing the objective function, a candidate key factor is eliminated when its regression coefficient shrinks to zero, thereby eliminating multicollinear variables.

5. The graph theory-based method for assessing the failure risk of human-machine coupling in shipping, as described in claim 2, is characterized in that... The step of using the causal random forest algorithm to mine the nonlinear causal relationship between the set of key risk factors and the machine failure data specifically includes: Multiple decision trees are constructed to form a forest structure, and each tree is trained based on a random subset of features; Machine failure is set as the dependent variable. The importance score of the variable is calculated by weighting the heterogeneity. Key risk factors with scores exceeding the preset first threshold and causal effect values ​​exceeding the preset second threshold are screened.

6. The graph theory-based method for assessing the failure risk of human-machine coupling in shipping, as described in claim 1, is characterized in that... The specific steps of constructing a weighted undirected network based on the set of key risk factors using graph theory include: Using each factor in the set of key risk factors as a node, and calculating the correlation strength between nodes based on standardized mutual information as the edge weight, a weighted undirected network is constructed.

7. The graph theory-based method for assessing the failure risk of human-machine coupling in shipping, as described in claim 1, is characterized in that... The steps for determining the priority of key risk factors specifically include: By calculating the shortest path distance of all node pairs in the time-varying weighted undirected network topology sequence, the global efficiency among risk factors is quantified using the global efficiency formula, and this is used as a benchmark value. Remove individual nodes and all their associated edges in turn to form subnetworks, and recalculate the global efficiency of each subnetwork. Based on the difference between the baseline value and the global efficiency of each sub-network, calculate the global efficiency decline rate for each node; The key risk factors are prioritized by sorting them in descending order of their global efficiency decline rate.

8. A graph theory-based system for assessing the risk of human-machine coupling failure in shipping, characterized in that, include: The extraction module is used to extract structured data, including timestamps, from the shipping accident database, and perform data cleaning and standardization to generate a valid dataset. The structured data includes risk factor data and machine failure data. The key factor generation module is used to filter the effective dataset and generate a set of key risk factors; The time-varying network generation module is used to construct a weighted undirected network based on the set of key risk factors using graph theory. It performs sliding segmentation on the effective dataset by setting a fixed-length time window and a preset step size, and recalculates the standardized mutual information values ​​between nodes and updates the edge weights within each time window, thereby generating a time-varying weighted undirected network topology sequence. The priority module is used to perform global efficiency analysis on the time-varying weighted undirected network topology sequence, calculate the global efficiency decline rate of the network after each node is removed, and sort the key risk factors according to the decline rate to determine the priority of the key risk factors. The output module is used to output corresponding accident prevention suggestions and crew training programs based on the priority of the key risk factors.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the graph theory-based shipping human-machine coupling failure risk assessment method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the graph theory-based shipping human-machine coupling failure risk assessment method according to any one of claims 1 to 7.

Images