A ship navigation risk prediction method and device, electronic equipment and medium

Abstract

The present application relates to a kind of ship navigation risk prediction method, device, electronic equipment and medium, belong to ship safety technical field, wherein, the method includes: the fault tree of ship capsizing is constructed;The fuzzy failure rate of each node in fault tree is calculated;The end value and the median of the range interval of index parameter risk interval table are fitted with correction coefficient, and the risk correction formula between the evaluation index parameter in interval and fuzzy failure rate is obtained;Data extraction is carried out in the range interval of index parameter by Latin hypercube sampling, and multiple sets of index data are obtained, multiple sets of index data are input into the risk correction formula obtained, and multiple sets of risk fuzzy failure rate are input into the preset formula to obtain multiple sets of risk probability data, to obtain sample data;The physical quantity data corresponding to target ship is input into the BP neural network model trained perfectly, and the risk probability value of target ship capsizing is obtained.The present application improves the precision of ship navigation risk prediction.

Description

Technical Field

[0001] This invention relates to the field of ship safety technology, and in particular to a method, device, electronic device and medium for predicting ship navigation risks. Background Technology

[0002] Ocean voyages are a vital pillar of global trade and economic development, and a primary mode of international energy, mineral, and food transportation. However, the operational environment of ocean voyages is highly uncertain, with ships potentially encountering various risks during transoceanic voyages, including severe weather, equipment failures, communication disruptions, and human error. Bulk carriers, as the mainstay of ocean voyages, are particularly prone to capsizing accidents due to their large cargo capacity, the liquefied and fluid nature of their cargo, and the complexity of ship handling. For example, the capsizing of the bauxite carrier "Bulk Jupiter" in Vietnam in 2015 and the bulk carrier "NurAllya" in Indonesia in 2019 resulted in more than 10 people missing and caused significant losses. Therefore, systematic modeling and analysis of the potential risks of bulk carriers are essential.

[0003] Fault Tree Analysis (FTA) is one of the most widely used methods in risk assessment. It can systematically explore the deep-seated causes of system failures by combining qualitative and quantitative methods, and its reliability has been verified in numerous engineering practices. Zhou Jinglong starts with the failure analysis of ship electrical equipment and discusses the application of FTA technology in ship electrical accidents. Rabiul Islam et al. used FTA to calculate the reliability of various subsystems and the overall main propulsion engine of a ship. However, in complex engineering systems such as ocean voyages, obtaining the accurate failure probability of numerous basic events is often difficult or even infeasible. To solve this problem, researchers have introduced fuzzy set theory to approximate the exact probability into a possibility, thus forming Fuzzy Fault Tree Analysis (FFTA), which can estimate the failure probability relatively accurately even under conditions of limited information. In recent years, FFTA has been widely developed and applied. For example, Chen and Hwang pioneered the method of converting linguistic variables into fuzzy numbers and developed fuzzy sorting to calculate failure rates; Celik et al. introduced FFTA into shipping accident investigations as a new risk modeling method. Sahin et al. extended fuzzy fault tree analysis by embedding an ontology-based fault tree structure to study collisions and groundings of autonomous vessels. Beyond this, applications include marine engineering, power systems, and energy systems. However, while existing research has made progress in many fields, it lacks a systematic risk prediction framework for real-world navigation risks in ocean-going vessels. FFTA can only achieve quantitative assessments under specific navigation parameters, and its analysis of the dynamic parameter evolution mapping of continuous navigation risks is insufficient, making it difficult to provide rapid and reliable support for real-time decision-making.

[0004] In summary, there is a lack of existing technologies for predicting ship navigation risks to improve the accuracy of such predictions. Summary of the Invention

[0005] In view of this, it is necessary to provide a method, device, electronic device and medium for predicting ship navigation risks, so as to solve the problem of low accuracy in ship navigation risk prediction in the prior art.

[0006] To address the aforementioned problems, in a first aspect, the present invention provides a method for predicting ship navigation risks, comprising:

[0007] Construct a fault tree for ship capsizing during navigation;

[0008] Calculate the fuzzy failure rate of each node in the fault tree;

[0009] The risk correction formula between the evaluation index parameters and the fuzzy failure rate of the fault tree nodes is obtained by fitting the end values ​​and median values ​​of the index parameter range in the preset index parameter risk interval table with the correction coefficient. The preset index parameter risk interval table includes: index parameters, index parameter range, and fuzzy failure rate correction coefficient. The median value is the average value of the end values ​​of the index parameter range interval, and the value in the middle of the index parameter range interval is the physical quantity corresponding to the node in the fault tree. The index parameters include: slight risk, low risk, moderate risk, high risk, and high risk. The evaluation index parameters within the interval are used to characterize the physical quantity corresponding to the ship capsizing event.

[0010] Data is extracted from the range of index parameters using Latin hypercube sampling to obtain multiple sets of index data. These multiple sets of index data are then input into the risk correction formula to obtain multiple sets of risk fuzzy failure rates. Finally, these multiple sets of risk fuzzy failure rates are input into the preset fault tree logic gate calculation formula to obtain multiple sets of ship capsizing event risk probability data.

[0011] Sample data was constructed based on multiple sets of indicator data and corresponding multiple sets of ship capsizing event risk probability data.

[0012] The physical quantity data corresponding to the target vessel is input into a fully trained BP neural network model to obtain the capsizing risk probability value of the target vessel. The capsizing risk probability value of the target vessel is used to predict the risk of the target vessel capsizing during navigation. The fully trained BP neural network model is trained based on sample data.

[0013] In one possible implementation, the root node of the fault tree is the ship capsizing accident, and the first leaf node under the root node includes environmental factors, human factors, and material factors. The second leaf node of the first leaf node corresponding to environmental factors includes visibility, wind conditions, and sea waves. The second leaf node of the first leaf node corresponding to human factors includes crew operational errors and crew decision-making ability. The second leaf node of the first leaf node corresponding to material factors includes cargo factors and equipment failure.

[0014] In one possible implementation, calculating the fuzzy failure rate of each node in the fault tree includes:

[0015] Based on fuzzy theory, linguistic variables are fuzzy processed through expert scoring to obtain the fuzzy failure probability of each node in the fault tree.

[0016] In one possible implementation, the failure probability obtained by fuzzy processing of linguistic variables through expert scoring based on fuzzy theory includes:

[0017] The evaluation language of experts is converted into fuzzy numbers by using membership functions, and the fuzzy numbers are aggregated to obtain aggregated fuzzy numbers.

[0018] The aggregated fuzzy number is converted into a fuzzy probability score based on the left-right fuzzy sorting method;

[0019] The fuzzy probability score is converted into the fuzzy failure rate of each event in the fault tree.

[0020] In one possible implementation, the expression for the fuzzy probability score is:

[0021]

[0022]

[0023]

[0024] In the formula, Let represent the fuzzy probability score, where h represents the first x-coordinate of the fuzzy membership function intersecting the x-axis, i represents the third x-coordinate of the fuzzy membership function intersecting y=1, j represents the fourth x-coordinate of the fuzzy membership function intersecting y=1, and k represents the second x-coordinate of the fuzzy membership function intersecting the x-axis. The x-coordinate represents the intersection of the fuzzy membership function and y = 1 - x. The x-coordinate represents the intersection of the fuzzy membership function and y=x.

[0025] In one possible implementation, the risk correction formula includes:

[0026]

[0027] In the formula, This represents the physical value corresponding to the overturning event. 'b' represents the first interval value of the indicator parameter, and 'b' represents the second interval value of the indicator parameter. express Correction coefficient for the corresponding fuzzy failure probability Indicates the right endpoint The corresponding correction coefficient for the fuzzy failure probability. Indicates the midpoint of the index parameter range ( + The fuzzy failure rate corresponding to ) / 2, This indicates a fuzzy failure rate.

[0028] In one possible implementation, the SHAP method is used to evaluate environmental, human, and physical factors to determine the factors that have the greatest impact on ship capsizing.

[0029] Secondly, the present invention also provides a ship navigation risk prediction device, comprising:

[0030] The fault tree construction module is used to construct fault trees for ship capsizing during navigation;

[0031] The fuzzy failure rate acquisition module is used to calculate the fuzzy failure rate of each node in the fault tree;

[0032] The fitting formula acquisition module is used to fit the endpoints and medians of the indicator parameter range in the preset indicator parameter risk interval table with the correction coefficient to obtain the risk correction formula between the evaluation indicator parameters within the interval and the fuzzy failure rate of the fault tree node. The preset indicator parameter risk interval table includes: indicator parameters, indicator parameter ranges, and fuzzy failure rate correction coefficients. The median value is the average of the endpoints of the indicator parameter range interval, and the median value is the physical quantity corresponding to the node in the fault tree. The indicator parameters include: slight risk, low risk, moderate risk, high risk, and high risk. The evaluation indicator parameters within the interval are used to characterize the physical quantity corresponding to the ship capsizing event.

[0033] The sample data acquisition module is used to extract data within the range of index parameters through Latin hypercube sampling to obtain multiple sets of index data. The multiple sets of index data are then input into the risk correction formula to obtain multiple sets of risk fuzzy failure rates. The multiple sets of risk fuzzy failure rates are then input into the preset fault tree logic gate calculation formula to obtain multiple sets of ship capsizing event risk probability data.

[0034] The sample data construction module is used to construct sample data based on multiple sets of indicator data and corresponding multiple sets of ship capsizing event risk probability data.

[0035] The capsizing risk prediction module is used to input the physical quantity data corresponding to the target vessel into a fully trained BP neural network model to obtain the capsizing risk probability value of the target vessel. The capsizing risk probability value of the target vessel is used to predict the risk of the target vessel capsizing during navigation. The fully trained BP neural network model is trained based on sample data.

[0036] Thirdly, the present invention also provides an electronic device, including a memory and a processor, wherein,

[0037] The memory is used to store programs;

[0038] The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the ship navigation risk prediction method described in any of the above implementations.

[0039] Fourthly, the present invention also provides a computer-readable storage medium for storing a computer-readable program or instructions, which, when executed by a processor, can implement the steps of the ship navigation risk prediction method described in any of the above implementations.

[0040] The beneficial effects of this invention are as follows: This invention provides a method for predicting ship navigation risks. This method includes constructing a fault tree for ship capsizing, systematically characterizing the hierarchical structure and logical relationships of risk factors during ocean-going vessel navigation, providing a reliable and comprehensive risk data foundation for subsequent neural network model training, calculating the fuzzy failure rate of each node in the fault tree, and fitting the endpoints and medians of the index parameter range intervals in a preset index parameter risk interval table with correction coefficients to obtain a risk correction formula between the evaluation index parameters within the interval and the fuzzy failure rate of the fault tree nodes. This risk correction formula can directly determine the ship capsizing risk value, greatly avoiding unnecessary calculations and improving computational efficiency. The preset index parameter risk interval table includes: index parameters, index parameter ranges, and fuzzy failure rate correction coefficients. The median value is the average of the endpoints of the parameter range intervals, and the values ​​of the index parameter range intervals are the nodes in the fault tree. The corresponding physical quantities, the index parameters include: slight risk, low risk, moderate risk, high risk, and high risk. The evaluation index parameters within the interval are used to characterize the physical quantities corresponding to the ship capsizing event. Data is extracted within the range of index parameters using Latin hypercube sampling to obtain multiple sets of index data. The multiple sets of index data are input into the risk correction formula to obtain multiple sets of risk fuzzy failure rates. The multiple sets of risk fuzzy failure rates are respectively input into the preset fault tree logic gate calculation formula to obtain multiple sets of ship capsizing event risk probability data. Sample data is constructed based on the multiple sets of index data and the corresponding multiple sets of ship capsizing event risk probability data. The physical quantity data corresponding to the target ship is input into the fully trained BP neural network model to obtain the target ship capsizing risk probability value. The target ship capsizing risk probability value is used to predict the risk of the target ship capsizing during navigation. The fully trained BP neural network model is trained based on the sample data. This invention uses fuzzy set theory to fuzzify the fault tree to obtain the fuzzy failure rate, and obtains a risk correction formula between the fuzzy failure rate and the physical quantity corresponding to the capsizing event through data fitting. It also combines a BP neural network to realize the systematic input of ship navigation parameters and dynamic autonomous risk assessment, thereby improving the accuracy of ship capsizing risk prediction. Attached Figure Description

[0041] Figure 1 A flowchart illustrating an embodiment of a ship navigation risk prediction method provided by the present invention;

[0042] Figure 2 This invention provides a method for predicting ship navigation risks, including a fault tree diagram of a bulk carrier capsizing during navigation.

[0043] Figure 3 This invention provides an embodiment of a ship navigation risk prediction method, which includes a membership function graph.

[0044] Figure 4 This invention provides an embodiment of a ship navigation risk prediction method, including fuzzy numbers and left / right score graphs.

[0045] Figure 5 This is a schematic diagram illustrating the failure rate correction principle of an embodiment of a ship navigation risk prediction method provided by the present invention.

[0046] Figure 6 A comparison chart of actual and predicted values ​​in a test set of a BP neural network model, which is an embodiment of the ship navigation risk prediction method provided by the present invention;

[0047] Figure 7 An example of a ship navigation risk prediction method provided by the present invention is shown in the error distribution and cumulative percentage curve of a BP neural network model prediction.

[0048] Figure 8 A feature importance analysis result diagram based on the SHAP method is provided as an embodiment of the ship navigation risk prediction method provided by the present invention;

[0049] Figure 9 A schematic flowchart of an embodiment of a ship navigation risk prediction device provided by the present invention;

[0050] Figure 10 A schematic diagram of an embodiment of the electronic device provided by the present invention. Detailed Implementation

[0051] 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 a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0052] In the description of the embodiments of the present invention, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0053] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

[0054] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0055] Before demonstrating the embodiments, the following terms will be explained.

[0056] Fuzzy set theory was proposed by Zadeh in 1965. Its core idea is to use membership functions to describe the "partial membership" relationship of elements to a set, thereby effectively characterizing uncertainty and fuzziness.

[0057] This invention provides a method, device, electronic device, and medium for predicting ship navigation risks, which will be described below.

[0058] Figure 1 This is a schematic flowchart of an embodiment of the ship navigation risk prediction method provided by the present invention, as shown below. Figure 1 As shown, the methods for predicting ship navigation risks include:

[0059] S101. Construct a fault tree for ship capsizing during navigation;

[0060] S102. Calculate the fuzzy failure rate of each node in the fault tree;

[0061] S103. Fit the endpoints and medians of the indicator parameter range in the preset indicator parameter risk interval table with the correction coefficient to obtain the risk correction formula between the evaluation indicator parameters within the interval and the fuzzy failure rate of the fault tree node. The preset indicator parameter risk interval table includes: indicator parameters, indicator parameter ranges, and fuzzy failure rate correction coefficients. The median value is the average of the endpoints of the indicator parameter range interval, and the median value is the physical quantity corresponding to the node in the fault tree. The indicator parameters include: slight risk, low risk, moderate risk, high risk, and high risk. The evaluation indicator parameters within the interval are used to characterize the physical quantity corresponding to the ship capsizing event.

[0062] S104. Data is extracted from the range of index parameters by Latin hypercube sampling to obtain multiple sets of index data. The multiple sets of index data are input into the risk correction formula to obtain multiple sets of risk fuzzy failure rates. The multiple sets of risk fuzzy failure rates are input into the preset fault tree logic gate calculation formula to obtain multiple sets of ship capsizing event risk probability data.

[0063] S105. Construct sample data based on multiple sets of indicator data and corresponding multiple sets of ship capsizing event risk probability data;

[0064] S106. Input the physical quantity data corresponding to the target vessel into the fully trained BP neural network model to obtain the capsizing risk probability value of the target vessel. The capsizing risk probability value of the target vessel is used to predict the risk of the target vessel capsizing during navigation. The fully trained BP neural network model is trained based on sample data.

[0065] Compared with existing technologies, this embodiment provides a method for predicting ship navigation risks. This method includes constructing a fault tree for ship capsizing, systematically characterizing the hierarchical structure and logical relationships of risk factors during ocean-going vessel navigation, providing a reliable and comprehensive risk data foundation for subsequent neural network model training, calculating the fuzzy failure rate of each node in the fault tree, and fitting the endpoints and medians of the index parameter range intervals in a preset index parameter risk interval table with correction coefficients to obtain a risk correction formula between the evaluation index parameters within the interval and the fuzzy failure rate of the fault tree nodes. This risk correction formula can directly determine the ship capsizing risk value, avoiding unnecessary calculations and improving computational efficiency. The preset index parameter risk interval table includes: index parameters, index parameter ranges, and fuzzy failure rate correction coefficients. The median value is the average of the endpoints of the index parameter range intervals, and the median value is the value of the node in the fault tree. The physical quantities corresponding to the points are index parameters including: slight risk, low risk, moderate risk, high risk, and high risk. The evaluation index parameters within the interval are used to characterize the physical quantities corresponding to the ship capsizing event. Data is extracted within the index parameter range using Latin hypercube sampling to obtain multiple sets of index data. These multiple sets of index data are input into the risk correction formula to obtain multiple sets of risk fuzzy failure rates. These multiple sets of risk fuzzy failure rates are then input into the preset fault tree logic gate calculation formula to obtain multiple sets of ship capsizing event risk probability data. Sample data is constructed based on the multiple sets of index data and the corresponding multiple sets of ship capsizing event risk probability data. The physical quantity data corresponding to the target ship is input into the fully trained BP neural network model to obtain the target ship capsizing risk probability value. The target ship capsizing risk probability value is used to predict the risk of the target ship capsizing during navigation. The fully trained BP neural network model is obtained based on the sample data. This invention uses fuzzy set theory to fuzzify the fault tree to obtain the fuzzy failure rate, and obtains a risk correction formula between the fuzzy failure rate and the physical quantity corresponding to the capsizing event through data fitting. It also combines a BP neural network to realize the systematic input of ship navigation parameters and dynamic autonomous risk assessment, thereby improving the accuracy of ship capsizing risk prediction.

[0066] It should be noted that this embodiment is applied to bulk carriers on ocean voyages. It is understood that it can be used for other types of ships that require assessment of capsizing risk.

[0067] In some embodiments of the present invention, the root node of the fault tree is a ship capsizing accident, and the first leaf node under the root node includes environmental factors, human factors and material factors. The second leaf node of the first leaf node corresponding to environmental factors includes visibility, wind conditions and sea waves. The second leaf node of the first leaf node corresponding to human factors includes crew operation errors and crew decision-making ability. The second leaf node of the first leaf node corresponding to material factors includes cargo factors and equipment failure.

[0068] In a specific embodiment of the present invention, to reduce the risk of capsizing during ocean-going bulk carrier voyages, it is necessary to identify various possible causes of capsizing risk and establish logical connections. Research revealed that the influencing factors mainly include three aspects: environmental factors, human factors, and material factors. Furthermore, more detailed basic events were identified through a layer-by-layer analysis. Combining expert experience, logic gates were assigned to the connections between different events. Table 1 shows the numbering explanation of the fault tree for ocean-going bulk carrier capsizing.

[0069] Table 1. Explanation of Fault Tree Numbers for Ocean-going Bulk Carrier Capsizing

[0070]

[0071] like Figure 2 The diagram shown is a fault tree diagram of a bulk carrier capsizing during navigation. The fault tree for ship capsizing during navigation is constructed based on the data in Table 1.

[0072] In some embodiments of the present invention, calculating the fuzzy failure rate of each node in the fault tree includes:

[0073] Based on fuzzy theory, the linguistic variables are fuzzy processed by expert scoring to obtain the failure probability of each node in the fault tree.

[0074] In some embodiments of the present invention, the fuzzy failure probability of each node in the fault tree obtained by performing fuzzy processing of linguistic variables through expert scoring based on fuzzy theory includes:

[0075] The evaluation language of experts is converted into fuzzy numbers by using membership functions, and the fuzzy numbers are aggregated to obtain aggregated fuzzy numbers.

[0076] The aggregated fuzzy number is converted into a fuzzy probability score based on the left-right fuzzy sorting method;

[0077] The fuzzy probability score is converted into the fuzzy failure rate of each node in the fault tree.

[0078] In a specific embodiment of the present invention, in risk assessment, many factors are often difficult to quantify with precise numerical values, and experts typically express their failure probabilities using linguistic variables. For example, seven fuzzy terms such as "very small," "small," "relatively small," "medium," "relatively large," "large," and "very large" can be represented by the symbols "VL," "L," "FL," "M," "FH," "H," and "VH," respectively. Each term corresponds to a membership function defined on the interval [0,1], used to reflect the confidence level of the failure probability of the basic event in different intervals, such as... Figure 3 The diagram shows the membership function. This method allows expert experience to be embedded quantitatively into the risk model, while the subsequent defuzzification process transforms the fuzzy language into concrete numerical values, thus providing computable data support for fault tree analysis and neural network modeling.

[0079] The failure probability of the basic event is determined by scores from different experts. To obtain a more objective single result, the scores from different experts need to be aggregated, as shown in formula (1):

[0080] (1)

[0081] in, and The number of representative experts and the number of basic events, w j These are the weighting coefficients for each expert. Membership function corresponding to expert scoring results. It is the aggregate fuzzy number of all basic events.

[0082] The fuzzy numbers are converted into fuzzy probability scores (FPS), which represent the clear values ​​of the basic event probabilities. The defuzzification process is performed based on the left-right fuzzy ranking method. The left and right functions are shown in formulas (2) and (3), respectively:

[0083] (2)

[0084] (3)

[0085] The functional relationships corresponding to fuzzy numbers intersect, forming, for example... Figure 4 The graph shown is an example of this. A, b, c, and d are the inflection points of the fuzzy membership function. and , intersecting the left and right sides of the fuzzy number function respectively. and Two points (Formula (4)). The fuzzy probability score (FPS) value of each basic event is obtained through Formula (5). In some embodiments of the present invention, the expression for the fuzzy probability score is:

[0086] ; (4)

[0087] (5)

[0088] In the formula, Let represent the fuzzy probability score, where h represents the first x-coordinate of the fuzzy membership function intersecting the x-axis, i represents the third x-coordinate of the fuzzy membership function intersecting y=1, j represents the fourth x-coordinate of the fuzzy membership function intersecting y=1, and k represents the second x-coordinate of the fuzzy membership function intersecting the x-axis. The x-coordinate represents the intersection of the fuzzy membership function and y = 1 - x. The x-coordinate represents the intersection of the fuzzy membership function and y=x.

[0089] All fuzzy probability scores (FPS) need to be converted into corresponding fuzzy failure rates (FPS). The relevant definitions are shown in formulas (6) and (7). In a specific embodiment of the present invention, the fuzzy failure probability corresponding to a ship capsizing event is defined.

[0090] (6)

[0091] (7)

[0092] In a specific embodiment of the present invention, based on a scoring standard using seven ambiguous languages ​​with different meanings (“VL”, “L”, “FL”, “M”, “FH”, “H”, and “VH”), 11 relevant experts were invited to score 36 indicators related to the navigation risks of ocean-going vessels. Some of the scoring results are shown in Table 2.

[0093] Table 2. Partial Results of Expert Scoring on Navigation Risk

[0094]

[0095] To ensure the assessment results were more accurate and reliable, several ocean-going captains were invited to score different basic events, with weights assigned based on the experts' experience levels. The scoring results established four primary indicators as standards for assessing the experts' experience: professional title, years of service, education level, and relevance of their major. Furthermore, multiple secondary indicators were developed, such as categorizing professional titles into captain, chief mate, second mate, and chief engineer, to obtain the risk awareness of personnel in different positions. The weighting coefficients of the 11 invited experts are shown in Table 3.

[0096] Table 3 Expert Information and Weighting of Navigation Risks for Bulk Carriers

[0097]

[0098] The process of obtaining the accurate failure probability includes the following steps: (1) obtaining the linguistic terminology used by different experts to evaluate the influencing factors of capsizing risk during ocean voyages; (2) aggregating the opinions of different experts to obtain an objective fuzzy number; (3) converting the fuzzy number into FPS; and (4) calculating the fuzzy failure rate. This paper uses X1 as an example to implement the above process, and the calculation flow is shown in Table 4. The calculation results of the fuzzy failure probability for all events are shown in Table 5.

[0099] Table 4. Calculation process of fuzzy failure probability

[0100]

[0101] Table 5 Calculation results of fuzzy failure probability

[0102]

[0103] In a specific embodiment of this invention, a suitable and reliable database must be obtained before constructing the BPNN model. However, existing data mostly originates from maritime accident investigation reports, which are limited by their non-standardized reporting format, resulting in limited data availability. Therefore, constructing a database based on virtual operating conditions using other methods becomes crucial. The designed virtual operating conditions should conform to real navigation conditions, and the range of all parameter values ​​should refer to relevant international conventions. Meanwhile, ocean voyage risks are influenced by numerous meteorological, human, and equipment parameters, and different parameter values ​​characterize the varying degrees of risk for bulk carrier ocean voyages. Therefore, it is necessary to empirically correct the impact ratio of different parameters when deviating from optimal conditions. The parameter ranges and correction coefficients for ocean voyage risk factors are shown in Table 6, and the data in Table 6 was obtained through questionnaire survey statistics. The failure probability correction standard was determined based on the comprehensive opinions of several front-line experts in ocean voyages, such as... Figure 5 As shown, the horizontal axis represents the possible values ​​of each index parameter during navigation, and the vertical axis represents the failure probability of different parameter values. By introducing a failure probability method based on virtual operating conditions and expert experience correction, the traditional FTA model is extended, transforming the originally static and discrete risk into a dynamic and continuous risk representation. This innovatively realizes dynamic risk modeling for bulk carrier ocean voyages under a complex multidimensional assessment system.

[0104] Table 6: Range of Parameters for Risk Factors in Bulk Carrier Navigation

[0105]

[0106] like Figure 5 The diagram shown illustrates the failure rate correction principle. Taking the general risk of X1 as an example, the index parameter values ​​range from [ab] to [1-2]. The probabilities corresponding to the extreme values ​​a, b, and the median (a+b) / 2 are respectively... , and , , The values ​​of c are 0.853, 1.147, and 1, respectively. This corresponds to the fuzzy failure probability calculated in formula (6). The failure probability for any value x within the range [a, b] As shown in formula (8), which is the risk correction formula, in some embodiments of the present invention, the risk correction formula includes:

[0107]

[0108] In the formula, This represents the physical value corresponding to the overturning event. 'b' represents the first interval value of the indicator parameter, and 'b' represents the second interval value of the indicator parameter. express Correction coefficient for the corresponding fuzzy failure probability Indicates the right endpoint The corresponding correction coefficient for the fuzzy failure probability. Indicates the midpoint of the index parameter range ( + The fuzzy failure rate corresponding to ) / 2, This indicates a fuzzy failure rate.

[0109] In a specific embodiment of the present invention, after obtaining the fuzzy failure rate, the probability value of the risk of capsizing during ocean voyages under the fault tree can be obtained by combining the minimal cut set (MCS) using formula (9).

[0110] (9)

[0111] In the formula, This represents the probability value indicating the risk of capsizing during navigation. This indicates a failure rate due to fuzzy target information.

[0112] By combining probability correction standards and actual navigation parameter ranges, an expanded database was obtained through Latin hypercube sampling. The corresponding capsizing risk probability value of ocean-going vessels was obtained through fault tree modeling, providing high-precision input and output sample data for the training of BP neural network. A total of 1000 sets of input and output data were obtained as shown in Table 7.

[0113] Table 7. Sampling of 1000 sets of bulk carrier index parameters

[0114]

[0115] To achieve optimal performance for BPNN, this embodiment systematically optimized key hyperparameters. The optimization process was based on three hyperparameters that significantly impact model performance: learning rate, number of hidden layer nodes, and dropout rate. The search space for the learning rate was set to [0.01, 0.001, 0.0001], covering the full range from high learning rates to fine-tuning; the number of hidden layer nodes was selected from five discrete values ​​[32, 64, 128, 256] to balance the model's expressive power and computational complexity; and the dropout rate was set to [0.1, 0.2, 0.3] to adjust the model's regularization strength. The dataset was divided into training and test sets in an 8:2 ratio, and the coefficient of determination was used as the primary evaluation metric for hyperparameter selection. By optimizing each parameter individually and performing multiple iterations, an optimal hyperparameter combination significantly superior to the baseline configuration was finally obtained, laying the foundation for subsequent model performance analysis and improved prediction accuracy.

[0116] Based on the above parameter settings, an optimal BPNN model was built. 800 sets of data were randomly selected as training data, and the remaining 200 sets were used as test data. The test results are as follows: Figure 6 and Figure 7 As shown. Figure 6 and Figure 7 The visualization results of BPNN predictions for ocean voyage risks are presented. Among them, Figure 6 This indicates that the coefficient of determination is as high as 0.9978 when comparing the actual values ​​and predicted values ​​in the test set. Furthermore, Figure 7 The distribution of prediction errors and their cumulative percentage curves are shown. As can be seen from the figure, the model prediction errors are mainly concentrated in the range of [−0.004, 0.003], exhibiting an approximately bell-shaped distribution, indicating that the overall prediction results are relatively stable. More than 80% of the sample prediction errors are controlled within the range of [−0.002, 0.002], indicating that the model can achieve high prediction accuracy in most cases.

[0117] In some embodiments of the present invention, the SHAP method is used to evaluate environmental factors, human factors, and physical factors in order to determine the factors that have the greatest impact on ship capsizing.

[0118] SHAP (SHapley Additive exPlanations) is a feature importance interpretation tool based on game theory, considered an effective method for evaluating the contribution of model input variables to prediction results. This method calculates the marginal contribution of each feature to the model output across all possible feature combinations, averages these contributions over all possible permutations, and finally obtains the SHAP value for each feature. This study uses the SHAP method to perform sensitivity analysis on the input variables of a BPNN model, obtaining the following results: Figure 8 The results are shown.

[0119] Figure 8 The results of feature importance analysis based on the SHAP method are presented. Overall, the SHAP values ​​of environmental factors are generally higher than those of human and material factors, indicating that environmental variables play a dominant role in the prediction of ocean voyage risks. Among environmental factors, features X10 (mean wave direction angle with bow), X2 (mean wind speed), X31 (cargo liquefaction), X11 (mean wind and wave direction angle with bow), and X7 (maximum wave height) have the highest SHAP values, indicating that these environmentally related indicators have the strongest sensitivity to model prediction results and are key driving factors affecting ocean voyage risks. In contrast, the SHAP values ​​of human factors (X18–X28) are generally low, indicating that their marginal contribution to model prediction is relatively limited, but still reflects the role of crew operation and management factors in risk control. Among material factors, the SHAP values ​​of X30, X31, and X32 are relatively prominent, showing that the status of the ship's cargo and the propulsion system have a significant impact on navigation risks. This result provides a reference for subsequent risk prevention and control, suggesting that priority should be given to environmental changes and the operating status of key equipment, while improving the overall safety level in conjunction with personnel management.

[0120] This embodiment constructs a capsizing risk prediction model for ocean-going bulk carriers based on FFTA and BPNN. FFTA introduces causal logic based on the real physical world into the model, which not only strengthens the expression of the risk mechanism but also significantly improves the interpretability of the prediction results. Furthermore, by combining Latin hypercube sampling and probability correction methods, a high-quality and representative virtual database is constructed, thus overcoming the limitation of limited actual accident data. Based on this database, the neural network can effectively learn the complex nonlinear mechanism of potential hazard factors on the capsizing risk of bulk carriers, achieving accurate capture of risk changes with dynamic data during navigation.

[0121] This embodiment innovatively constructs a capsizing risk prediction model for ocean-going bulk carriers based on FFTA and BPNN. FFTA introduces causal logic based on the real physical world into the model, which not only strengthens the expression of the risk mechanism but also significantly improves the interpretability of the prediction results. Furthermore, by combining Latin hypercube sampling and probability correction methods, a virtual database that conforms to real-world navigation conditions is constructed, overcoming the limitation of limited actual accident data. Based on this database, the neural network can effectively learn the complex nonlinear mechanism of potential hazard factors on the capsizing risk of bulk carriers, achieving accurate capture of risk changes with dynamic data during navigation. The main conclusions of this embodiment are as follows:

[0122] (1) Based on actual navigation conditions and expert experience, a fault tree model was built, which includes three major categories of environment, people and objects and 36 basic events. This model systematically represents the hierarchical structure and logical relationship of risk factors during the navigation of ocean bulk carriers, providing a reliable and comprehensive risk data foundation for the training of subsequent neural network models.

[0123] (2) Based on expert experience, risk correction functions were established for different degrees of deviation of ocean navigation parameters, realizing the effective combination of FFTA risk system and BPNN prediction model for ocean bulk carrier navigation.

[0124] (3) By optimizing the neural network structure and training parameters, a prediction model for ocean-going bulk carrier navigation risk with both fast response and high prediction accuracy (with a determination coefficient as high as 0.9978) was established. This model not only significantly improved the fitting performance and generalization ability of the model, but also met the dual requirements of real-time performance and accuracy in the ocean-going navigation scenario of bulk carriers.

[0125] (4) The results of the sensitivity analysis revealed the main factors affecting the navigation risk of ocean-going bulk carriers, including the mean wave direction and bow angle, mean wind speed, cargo liquefaction, mean wind and wave direction and bow angle and maximum wave height. Moreover, the importance of environmental factors is far greater than that of human factors and material factors. It is necessary to pay attention to the occurrence of adverse environments in actual navigation.

[0126] Implementing a ship navigation risk prediction method according to an embodiment of the present invention, based on a ship navigation risk prediction method, correspondingly, such as... Figure 9 As shown, this embodiment of the invention also provides a ship navigation risk prediction method apparatus. A ship navigation risk prediction method apparatus 900 includes:

[0127] Fault tree construction module 901 is used to construct fault trees for ship capsizing during navigation;

[0128] The fuzzy failure rate acquisition module 902 is used to calculate the fuzzy failure rate of each node in the fault tree;

[0129] The fitting formula acquisition module 903 is used to fit the endpoints and medians of the index parameter range in the preset index parameter risk interval table with the correction coefficient to obtain the risk correction formula between the evaluation index parameter within the interval and the fuzzy failure rate of the fault tree node. The preset index parameter risk interval table includes: index parameter, index parameter range and fuzzy failure rate correction coefficient. The median value is the average value of the endpoints of the index parameter range interval, and the value in the middle of the index parameter range interval is the physical quantity corresponding to the node in the fault tree. The index parameter includes: slight risk, low risk, moderate risk, high risk and high risk. The evaluation index parameter within the interval is used to characterize the physical quantity corresponding to the ship capsizing event.

[0130] The sample data acquisition module 904 is used to extract data in the range of index parameters through Latin hypercube sampling to obtain multiple sets of index data. The multiple sets of index data are input into the risk correction formula to obtain multiple sets of risk fuzzy failure rates. The multiple sets of risk fuzzy failure rates are respectively input into the preset fault tree logic gate calculation formula to obtain multiple sets of ship capsizing event risk probability data.

[0131] The sample data construction module 905 is used to construct sample data based on multiple sets of indicator data and corresponding multiple sets of ship capsizing event risk probability data.

[0132] The capsizing risk prediction module 906 is used to input the physical quantity data corresponding to the target vessel into a fully trained BP neural network model to obtain the capsizing risk probability value of the target vessel. The capsizing risk probability value of the target vessel is used to predict the risk of the target vessel capsizing during navigation. The fully trained BP neural network model is trained based on sample data.

[0133] The ship navigation risk prediction device 900 provided in the above embodiments can realize the technical solution described in the above embodiment of the ship navigation risk prediction method. The specific implementation principle of each module or unit can be found in the corresponding content in the above embodiment of the ship navigation risk prediction method, and will not be repeated here.

[0134] like Figure 10 As shown, the present invention also provides an electronic device 1000. The electronic device 1000 includes a processor 1001, a memory 1002, and a display 1003. Figure 10 Only some components of the electronic device 1000 are shown, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

[0135] In some embodiments, processor 1001 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 1002 or process data, such as a ship navigation risk prediction method in this invention.

[0136] In some embodiments, processor 1001 may be a single server or a group of servers. The server group may be centralized or distributed. In some embodiments, processor 1001 may be local or remote. In some embodiments, processor 1001 may be implemented on a cloud platform. In some embodiments, the cloud platform may include a private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, internal cloud, multi-cloud, or any combination thereof.

[0137] In some embodiments, memory 1002 may be an internal storage unit of electronic device 300, such as a hard disk or memory of electronic device 1000. In other embodiments, memory 1002 may also be an external storage device of electronic device 1000, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on electronic device 1000.

[0138] Furthermore, the memory 1002 may include both internal storage units of the electronic device 1000 and external storage devices. The memory 1002 is used to store application software and various types of data installed on the electronic device 1000.

[0139] In some embodiments, display 1003 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 1003 is used to display information from electronic device 1000 and to display a visual user interface. Components 1001-1003 of electronic device 1000 communicate with each other via a system bus.

[0140] In one embodiment, when the processor 1001 executes a ship navigation risk prediction method program stored in the memory 1002, the following steps can be implemented:

[0141] Construct a fault tree for ship capsizing during navigation;

[0142] Calculate the fuzzy failure rate of each node in the fault tree;

[0143] The risk correction formula between the evaluation index parameters and the fuzzy failure rate of the fault tree nodes is obtained by fitting the end values ​​and median values ​​of the index parameter range in the preset index parameter risk interval table with the correction coefficient. The preset index parameter risk interval table includes: index parameters, index parameter range, and fuzzy failure rate correction coefficient. The median value is the average value of the end values ​​of the index parameter range interval, and the value in the middle of the index parameter range interval is the physical quantity corresponding to the node in the fault tree. The index parameters include: slight risk, low risk, moderate risk, high risk, and high risk. The evaluation index parameters within the interval are used to characterize the physical quantity corresponding to the ship capsizing event.

[0144] Data is extracted from the range of index parameters using Latin hypercube sampling to obtain multiple sets of index data. These multiple sets of index data are then input into the risk correction formula to obtain multiple sets of risk fuzzy failure rates. Finally, these multiple sets of risk fuzzy failure rates are input into the preset fault tree logic gate calculation formula to obtain multiple sets of ship capsizing event risk probability data.

[0145] Sample data was constructed based on multiple sets of indicator data and corresponding multiple sets of ship capsizing event risk probability data.

[0146] The physical quantity data corresponding to the target vessel is input into a fully trained BP neural network model to obtain the capsizing risk probability value of the target vessel. The capsizing risk probability value of the target vessel is used to predict the risk of the target vessel capsizing during navigation. The fully trained BP neural network model is trained based on sample data.

[0147] It should be understood that when the processor 1001 executes a ship navigation risk prediction method program in the memory 1002, in addition to the functions mentioned above, it can also perform other functions, as can be found in the description of the corresponding method embodiments above.

[0148] Furthermore, the embodiments of the present invention do not specifically limit the type of the electronic device 1000 mentioned. The electronic device 1000 can be a mobile phone, tablet computer, personal digital assistant (PDA), wearable device, laptop computer, or other portable electronic device. Exemplary embodiments of portable electronic devices include, but are not limited to, portable electronic devices running iOS, Android, Microsoft, or other operating systems. The aforementioned portable electronic device can also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the present invention, the electronic device 1000 may not be a portable electronic device, but rather a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0149] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0150] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting ship navigation risks, characterized in that, include: Construct a fault tree for ship capsizing during navigation; Calculate the fuzzy failure rate of each node in the fault tree; The risk correction formula between the evaluation index parameters and the fuzzy failure rate of the fault tree nodes is obtained by fitting the end values ​​and median values ​​of the index parameter range in the preset index parameter risk interval table with the correction coefficient. The preset index parameter risk interval table includes: index parameters, index parameter ranges and correction coefficients. The median value is the average value of the end values ​​of the index parameter range interval, and the value in the middle of the index parameter range interval is the physical quantity corresponding to the node in the fault tree. The index parameters include: slight risk, low risk, moderate risk, high risk and high risk. The evaluation index parameters in the interval are used to characterize the physical quantity corresponding to the ship capsizing event. Data is extracted from the range of index parameters by Latin hypercube sampling to obtain multiple sets of index data. These multiple sets of index data are then input into the risk correction formula to obtain multiple sets of risk fuzzy failure rates. Finally, these multiple sets of risk fuzzy failure rates are input into the preset fault tree logic gate calculation formula to obtain multiple sets of ship capsizing event risk probability data. Sample data was constructed based on multiple sets of indicator data and corresponding multiple sets of ship capsizing event risk probability data. The physical quantity data corresponding to the target vessel is input into a fully trained BP neural network model to obtain the capsizing risk probability value of the target vessel. The capsizing risk probability value of the target vessel is used to predict the risk of the target vessel capsizing during navigation. The fully trained BP neural network model is trained based on sample data.

2. The ship navigation risk prediction method according to claim 1, characterized in that, The root node of the fault tree is the ship capsizing accident. The first leaf node under the root node includes environmental factors, human factors, and material factors. The second leaf node of the first leaf node corresponding to environmental factors includes visibility, wind conditions, and sea waves. The second leaf node of the first leaf node corresponding to human factors includes crew operational errors and crew decision-making ability. The second leaf node of the first leaf node corresponding to material factors includes cargo factors and equipment failure.

3. The ship navigation risk prediction method according to claim 1, characterized in that, The calculation of the fuzzy failure rate of each node in the fault tree includes: Based on fuzzy theory, linguistic variables are fuzzy processed through expert scoring to obtain the fuzzy failure probability of each node in the fault tree.

4. The ship navigation risk prediction method according to claim 3, characterized in that, The method of fuzzy processing of linguistic variables based on fuzzy theory and expert scoring to obtain the failure probability of each node in the fault tree includes: The evaluation language of experts is converted into fuzzy numbers by using membership functions, and the fuzzy numbers are aggregated to obtain aggregated fuzzy numbers. The aggregated fuzzy number is converted into a fuzzy probability score based on the left-right fuzzy sorting method; The fuzzy probability score is converted into the fuzzy failure rate of each node in the fault tree.

5. The ship navigation risk prediction method according to claim 4, characterized in that, The expression for the fuzzy probability score is: In the formula, Let represent the fuzzy probability score, where h represents the first x-coordinate of the fuzzy membership function intersecting the x-axis, i represents the third x-coordinate of the fuzzy membership function intersecting y=1, j represents the fourth x-coordinate of the fuzzy membership function intersecting y=1, and k represents the second x-coordinate of the fuzzy membership function intersecting the x-axis. The x-coordinate represents the intersection of the fuzzy membership function and y = 1 - x. The x-coordinate represents the intersection of the fuzzy membership function and y=x.

6. The ship navigation risk prediction method according to claim 1, characterized in that, The risk correction formula includes: In the formula, This represents the physical value corresponding to the overturning event. 'b' represents the first interval value of the indicator parameter, and 'b' represents the second interval value of the indicator parameter. express The corresponding correction coefficient for the fuzzy failure probability. Indicates the right endpoint The corresponding correction coefficient for the fuzzy failure probability. Indicates the midpoint of the index parameter range ( + The fuzzy failure rate corresponding to ) / 2, This indicates a fuzzy failure rate.

7. The method for predicting ship navigation risks according to claim 1, characterized in that, The SHAP method is used to evaluate environmental, human, and physical factors in order to determine the factors that have the greatest impact on ship capsizing.

8. A ship navigation risk prediction device, characterized in that, include: The fault tree construction module is used to construct fault trees for ship capsizing during navigation; The fuzzy failure rate acquisition module is used to calculate the fuzzy failure rate of each node in the fault tree; The fitting formula acquisition module is used to fit the endpoints and medians of the indicator parameter range in the preset indicator parameter risk interval table with the correction coefficient to obtain the risk correction formula between the evaluation indicator parameters within the interval and the fuzzy failure rate of the fault tree node. The preset indicator parameter risk interval table includes: indicator parameters, indicator parameter ranges, and fuzzy failure rate correction coefficients. The median value is the average of the endpoints of the indicator parameter range interval, and the median value is the physical quantity corresponding to the node in the fault tree. The indicator parameters include: slight risk, low risk, moderate risk, high risk, and high risk. The evaluation indicator parameters within the interval are used to characterize the physical quantity corresponding to the ship capsizing event. The sample data acquisition module is used to extract data within the range of index parameters through Latin hypercube sampling to obtain multiple sets of index data. The multiple sets of index data are then input into the risk correction formula to obtain multiple sets of risk fuzzy failure rates. The multiple sets of risk fuzzy failure rates are then input into the preset fault tree logic gate calculation formula to obtain multiple sets of ship capsizing event risk probability data. The sample data construction module is used to construct sample data based on multiple sets of indicator data and corresponding multiple sets of ship capsizing event risk probability data. The capsizing risk prediction module is used to input the physical quantity data corresponding to the target vessel into a fully trained BP neural network model to obtain the capsizing risk probability value of the target vessel. The capsizing risk probability value of the target vessel is used to predict the risk of the target vessel capsizing during navigation. The fully trained BP neural network model is trained based on sample data.

9. An electronic device, characterized in that, Including memory and processor, among which, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the ship navigation risk prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the ship navigation risk prediction method according to any one of claims 1 to 7.

Images