Ship accident risk assessment method and system suitable for arctic water area

Abstract

The invention relates to a ship accident risk assessment method and system suitable for an arctic water area. The method comprises the following steps: acquiring static attribute information and navigation plan information of a target ship; dividing the planned voyage into a plurality of time-space points, and determining a risk scene to which each time-space point belongs; according to the environmental data of each time-space point, determining a key influence factor corresponding to the risk scene to which the environmental data belongs through a bivariate Probit model; acquiring the navigation behavior characteristics of the target ship on the planned voyage of the target ship; taking the navigation behavior characteristics and the key influence factors as input, and obtaining the joint probability of different types of accident consequences of the target ship at each time-space point through a scene-divided accident cause model; and according to the joint probability, quantifying a risk value of the target ship at each time-space point in the planned voyage, and carrying out risk grade division according to a preset threshold to obtain an accident risk assessment result. Compared with the prior art, the accuracy of accident risk assessment and the adaptability to the north pole complex environment are remarkably improved.

Description

Technical Field

[0001] This invention relates to the intersection of maritime traffic safety and intelligent data analysis, and in particular to a method and system for assessing ship accident risks in Arctic waters. Background Technology

[0002] As global warming accelerates the melting of Arctic sea ice, it drives the development of commercial shipping in the Arctic shipping routes. However, the extreme environment in the Arctic waters makes the risk of ship accidents much higher than in conventional sea areas, and once an accident occurs, it can easily lead to serious casualties, property damage and environmental pollution.

[0003] While preliminary technologies for ship navigation safety in Arctic waters have been developed based on discrete choice models and Bayesian networks for analyzing influencing factors, there are still technical shortcomings in multi-source heterogeneous data processing and scenario adaptability. First, the structure of Arctic ship accident data differs greatly from that of environmental data such as sea ice and meteorology. Existing fusion methods do not take into account the spatiotemporal dynamics of high-latitude Arctic data, resulting in low data matching accuracy. Secondly, existing technologies mostly use global models to analyze influencing factors, without modeling specific scenarios based on the significant seasonal and latitudinal differences in the Arctic, resulting in insufficient accuracy in identifying key factors. Third, the coupling relationship between the severe consequences of the accident and pollution is not adequately considered, making it difficult to fully reflect the complex risk characteristics of Arctic accidents.

[0004] For example, patent application CN119360677A discloses a method and system for risk management of polar vessel navigation. This method uses a random forest algorithm to interpolate and fuse multi-source environmental spatiotemporal data such as polar wind speed and sea ice, constructs a Bayesian network to assess the risks of accidents such as ice entrapment and collisions, and combines vessel seaworthiness, navigation windows, routes, and control measures to form a four-dimensional decision-making process. Although this method achieves the fusion processing of environmental data, it does not establish a heterogeneous coupling mechanism between vessel accident data and environmental data such as sea ice and meteorology. Furthermore, it adopts a global static model framework and does not dynamically model different scenarios for the significant seasonal-latitude differences in the Arctic. At the same time, it lacks a quantitative assessment dimension of environmental pollution consequences, resulting in insufficient accuracy in identifying key risk factors and difficulty in fully reflecting the coupled characteristics of the severe and pollution consequences of Arctic accidents. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a method and system for ship accident risk assessment applicable to Arctic waters, which significantly improves the accuracy of accident risk assessment and adaptability to the complex Arctic environment.

[0006] The objective of this invention can be achieved through the following technical solutions: A method for assessing ship accident risks applicable to Arctic waters includes the following steps: Obtain the static attribute information and navigation plan information of the target vessel; Based on the navigation plan information, the target vessel's planned voyage is divided into multiple spatiotemporal points, and the risk scenario associated with each spatiotemporal point is determined. Based on the environmental data at each spatiotemporal point, the key influencing factors corresponding to the risk scenario to which it belongs are determined by a bivariate Probit model. Based on the static attribute information and the navigation plan information, the navigation behavior characteristics of the target vessel on its planned voyage are obtained; The navigation behavior characteristics of the target vessel on its planned voyage and the key influencing factors corresponding to the risk scenario at each time and space point are used as inputs. The joint probability of the target vessel experiencing different types of accident consequences at each time and space point is obtained through the sub-scenario accident causation model. Based on the combined probability of different types of accident consequences occurring at various time and space points for the target vessel, the risk value of the target vessel at various time and space points in the planned voyage is quantified, and the risk level is classified according to the preset threshold to obtain the accident risk assessment result.

[0007] Furthermore, the static attribute information includes the ship's IMO number, flag state, ship type, gross tonnage, and age, while the voyage plan information includes the planned route, planned speed, and start and end times of the voyage.

[0008] Furthermore, the risk scenarios are divided according to seasonal and latitudinal dimensions, including summer-high latitude scenarios, summer-low latitude scenarios, non-summer-high latitude scenarios, and non-summer-low latitude scenarios.

[0009] Furthermore, the environmental data includes sea ice data and meteorological data. The sea ice data includes sea ice concentration and sea ice thickness, and the meteorological data includes wind speed data at a height of 10 meters above the sea surface.

[0010] Furthermore, the navigation behavior characteristics include accident type tendency characteristics, route key point operation characteristics, and spatiotemporal exposure characteristics. The accident type tendency characteristics are the historical probability distribution of various accidents occurring in the same risk scenario for ships with the same static attribute information as the target ship. The route key point operation characteristics are the preset navigation operations of the target ship in response to high-risk sections on the planned route. The spatiotemporal exposure characteristics are the exposure duration and path of the target ship in different risk scenarios.

[0011] Furthermore, the bivariate Probit model, using severe accidents and pollution accidents as dependent variables and the final explanatory variable system as independent variables, determines the key influencing factors corresponding to the risk scenario at each spatiotemporal point using the bivariate Probit model through the following specific steps: Maximum likelihood estimation is performed on bivariate Probit models under different risk scenarios, outputting the coefficients, standard deviations, Z-statistics, and probability values ​​of each explanatory variable. Significant variables under each risk scenario are then identified, revealing the influencing factors. The bivariate Probit model is as follows: In the formula, As a latent variable for the severity of serious accidents, To and The corresponding set of explanatory variables, As a latent variable representing the degree of pollution in a pollution incident, To and The corresponding set of explanatory variables, For the regression coefficient vector, For the disturbance term, It is a joint standard normal distribution. For disturbance terms The mean, For disturbance terms The mean, For disturbance terms variance For disturbance terms variance The correlation coefficient between serious accidents and pollution accidents; The serious accident is as follows: The pollution incident was as follows: In the formula, This is a serious accident. For pollution incident; Calculate the marginal effect of each influencing factor on the probability of the simultaneous occurrence of serious accidents and pollution accidents, and sort the influencing factors according to the magnitude of the absolute value of the marginal effect to screen out the key influencing factors.

[0012] Furthermore, the formula for calculating the marginal effect of each influencing factor on the probability of the simultaneous occurrence of serious accidents and pollution accidents is as follows: In the formula, For the first One influencing factor The marginal effect of the probability of a serious accident and a pollution accident occurring simultaneously. For the set of influencing factors, In the set of influencing factors The probability of serious accidents and pollution accidents occurring simultaneously under certain conditions. Let be the probability density function of a bivariate normal distribution. For the first One influencing factor The direct impact coefficient on serious accidents. For the first One influencing factor The direct impact coefficient on pollution accidents.

[0013] Furthermore, the joint probability of different types of accident consequences occurring at various spatiotemporal points is: In the formula, The joint probability of different types of accident consequences occurring at various spatiotemporal points. This is a serious accident. For pollution incidents, For the first One risk scenario, As a key influencing factor, For navigation behavior characteristics, For normalization function, For risk scenarios The weight parameters of the output layer, For activation function, For risk scenarios The weight parameters of the hidden layer, For splicing operations, For risk scenarios The bias parameters of the hidden layer, For risk scenarios The bias parameters of the output layer.

[0014] Furthermore, the joint probabilities of different types of accident consequences occurring at various spatiotemporal points include the first joint probability that the target vessel has a serious accident and a pollution accident at the spatiotemporal point, the second joint probability that it has a serious accident but no pollution accident, and the third joint probability that it has no serious accident but a pollution accident.

[0015] Furthermore, the risk values ​​of the target vessel at various spatiotemporal points during its planned voyage are: In the formula, For the target vessel during the planned voyage Risk value at each spatiotemporal point For the target vessel during the planned voyage The first joint probability of each spatiotemporal point The first joint probability weight, For the target vessel during the planned voyage The second joint probability of each spatiotemporal point For the second joint probability weight, For the target vessel during the planned voyage The third joint probability of each spatiotemporal point The third joint probability weight; The accident risk assessment results are as follows: In the formula, The low-risk threshold The threshold for medium risk. This is a high-risk threshold.

[0016] According to another aspect of the present invention, a ship accident risk assessment system suitable for Arctic waters is provided, comprising: The information acquisition module is used to acquire static attribute information and navigation plan information of the target vessel; The spatiotemporal division module is used to divide the planned voyage of the target vessel into multiple spatiotemporal points based on the navigation plan information, and to determine the risk scenario to which each spatiotemporal point belongs. The key influencing factor identification module is used to determine the key influencing factors corresponding to the risk scenario to which each spatiotemporal point belongs by using a bivariate Probit model based on the environmental data of each spatiotemporal point. The navigation behavior feature acquisition module is used to obtain the navigation behavior features of the target vessel on its planned voyage based on the static attribute information and the navigation plan information. The joint probability acquisition module is used to take the key influencing factors and navigation behavior characteristics corresponding to the risk scenario of each spatiotemporal point as input, and obtain the joint probability of the target ship experiencing different types of accident consequences at each spatiotemporal point through the sub-scenario accident causation model. The accident risk assessment module is used to quantify the risk value of the target vessel at each time and space point in the planned voyage based on the joint probability of different types of accident consequences occurring at each time and space point, and to classify the risk level according to the preset threshold to obtain the accident risk assessment result.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention employs a bivariate Probit model, utilizing maximum likelihood estimation to screen significant variables and calculate the marginal effect of joint probabilities. This screens out key influencing factors corresponding to the risk scenario at each spatiotemporal point. Based on static attribute information and navigation plan information, it obtains the navigation behavior characteristics of the target vessel on its planned voyage. Through a scenario-based accident causation model, combined with Softmax function normalization, it outputs the joint probability of severe accidents and pollution accidents. This achieves accurate identification of key influencing factors under different risk scenarios and quantitative analysis of the coupling relationship between the two consequences of accidents. It overcomes the limitations of existing technologies, such as insufficient generalization of global modeling and neglect of consequence correlation, resulting in weak early warning capabilities and poor practicality. This significantly improves the adaptability of the risk assessment model to the complex Arctic environment and the accuracy of accident risk assessment.

[0018] 2. This invention generates spatiotemporal points by interpolation at hourly intervals based on navigation plan information, and dynamically determines each spatiotemporal point by combining seasonal and latitude dimensions. Based on the environmental data of each spatiotemporal point, the key influencing factors corresponding to its respective risk scenario are determined through a bivariate Probit model. The navigation behavior characteristics of the target vessel on its planned voyage and the key influencing factors corresponding to the risk scenario of each spatiotemporal point are used as inputs to the sub-scenario accident causation model. This achieves high-precision spatiotemporal matching and collaborative analysis of ship data, accident data, and environmental data, and solves the problems of low fusion efficiency and large matching errors caused by the failure to consider the spatiotemporal dynamics of data in existing Arctic risk assessment technologies. This significantly improves the data foundation reliability and robustness of the risk assessment model. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating a ship accident risk assessment method applicable to Arctic waters proposed in this invention. Figure 2 This is a schematic diagram of a ship accident risk assessment system applicable to Arctic waters proposed in this invention. Detailed Implementation

[0020] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0021] Example 1 This embodiment provides a method for assessing ship accident risks applicable to Arctic waters, such as... Figure 1 As shown, it includes the following steps: S1. Obtain the static attribute information and navigation plan information of the target vessel.

[0022] Static attribute information includes the vessel's IMO number, flag state, vessel type, gross tonnage, and age. Navigation plan information includes the planned route, planned speed, and start and end times of the voyage.

[0023] S2. Divide the target vessel's planned voyage into multiple time and space points based on the voyage plan information, and determine the risk scenario to which each time and space point belongs.

[0024] A spatiotemporal point sequence is generated based on the target vessel's navigation plan. The navigation plan includes a series of waypoints (latitude and longitude coordinates) and estimated arrival times.

[0025] Interpolation is performed between adjacent waypoints at preset time intervals (e.g., 1 hour) to calculate the precise latitude and longitude coordinates of each interval point, capturing the continuous changes in environmental conditions during the voyage.

[0026] Each calculated location is uniquely defined by a timestamp (accurate to the hour) and geographic coordinates (latitude and longitude), forming a spatiotemporal point. The entire planned flight path is transformed into an ordered set of spatiotemporal points.

[0027] The risk scenarios for each spatiotemporal point are divided according to seasonal and latitudinal dimensions, including summer-high latitude scenarios, summer-low latitude scenarios, non-summer-high latitude scenarios, and non-summer-low latitude scenarios.

[0028] Seasonal division: Summer scenario: When the date of the space-time point falls between May 1st and September 30th, it is classified as the summer scenario. During this period, Arctic sea ice melts significantly, shipping lanes open, and shipping activities are frequent.

[0029] Non-summer scenario: When the date of the space-time point falls between October 1st and April 30th of the following year, it is classified as a non-summer scenario. During this period, sea ice grows rapidly or is at its peak, the environment is harsh, and navigation is difficult.

[0030] Latitude and longitude division: High-latitude scenario: When the latitude coordinates of a point in space are greater than or equal to 66°34′N, it is classified as a high-latitude scenario. This region experiences significant polar day and polar night phenomena, and sea ice conditions are complex.

[0031] Low-latitude scenario: When the latitude coordinates of a time point are less than 66°34′N, it is classified as a low-latitude scenario. Although this area is within the Arctic waters, the environmental conditions are relatively mild.

[0032] S3. Based on the environmental data at each spatiotemporal point, the key influencing factors corresponding to the risk scenario to which it belongs are determined through a bivariate Probit model.

[0033] Environmental data includes sea ice data and meteorological data. Sea ice data includes sea ice concentration and sea ice thickness, while meteorological data includes wind speed data at a height of 10 meters above the sea surface.

[0034] The bivariate Probit model uses severe accidents and pollution accidents as dependent variables and the final explanatory variable system as independent variables. Its construction steps include: A multi-source heterogeneous data source system for Arctic waters was constructed to acquire basic data on ship accidents, sea ice environment data, and meteorological data. The basic ship accident data includes ship attributes, accident attributes, and accident consequence attributes. The sea ice environment data includes sea ice concentration and thickness. The meteorological data includes wind speed data at 10 meters above sea level. The basic ship accident data includes global ship accident records, with core fields including ship IMO number, flag state, ship type, gross tonnage, ship age, accident time, latitude and longitude, accident type, number of casualties, property damage, and pollution leakage. The sea ice environment data includes daily sea ice concentration data (1 km resolution, 0-100%) based on MODIS and AMSR2 sensors and daily sea ice thickness data (40 km resolution, 0-51 cm range) based on SMOS and SMAP sensors, in NetCDF format. The meteorological data includes wind speed data at 10 meters above sea level for the corresponding area at the time of the accident, with a temporal resolution of 1 hour and a spatial resolution of 0.25°.

[0035] Preprocessing of multi-source heterogeneous data includes data cleaning, spatial filtering and spatiotemporal alignment, deletion of abnormal accident data and invalid environmental data, filtering of target area data based on the range of Arctic waters, and achieving accurate spatiotemporal matching of accident data and environmental data based on accident timestamps and latitude and longitude.

[0036] The preprocessing process is as follows: Data cleaning involved removing records from ship accident data that were older than 50 years, had no accident type, or had missing latitude and longitude coordinates or were located outside the Arctic waters (50°05′-82°00′N, 159°00′E-172°12′W), and correcting contradictory data regarding "total loss accidents without serious consequences." For sea ice environmental data, outliers with sea ice concentration <0 or >100% and sea ice thickness <0 or >51cm were removed, and the average of adjacent grid cells was used to fill in missing daily data. For meteorological data, linear interpolation was used to supplement missing wind speed values ​​within one hour, and meteorological records missing for more than six consecutive hours were deleted.

[0037] Spatial filtering involves converting the latitude and longitude coordinates of ship accident data and environmental data to the WGS84 coordinate system based on the Arctic waters vector boundary map. The "spatial connectivity" tool is then used to filter data whose latitude and longitude fall within the Arctic waters boundary, eliminating redundant records from non-Arctic regions.

[0038] For spatiotemporal alignment, the timestamps of ship accident data are uniformly converted to UTC time (accurate to the hour). Sea ice environmental data are aggregated into daily averages according to the natural day in which the accident occurred. Meteorological data directly extracts the wind speed data corresponding to the accident hour. Based on the latitude and longitude of the accident, the grid points of sea ice and meteorological data are matched using the nearest neighbor interpolation method to ensure that the spatial matching error is ≤5km, and finally a preliminary dataset with no missing data and spatiotemporal consistency is formed.

[0039] A hierarchical fusion strategy was adopted to fuse the preprocessed multi-source heterogeneous data and construct a three-dimensional dataset of "ship-accident-environment". The ship dimension includes flag state, ship type, gross tonnage and ship age. The accident dimension includes accident type and accident time. The environment dimension includes sea ice concentration, sea ice thickness and wind speed.

[0040] Using "IMO number + accident timestamp" as the unique association key, a mapping relationship was established between basic ship accident data and sea ice environmental data and meteorological data. For missing ship gross tonnage and age fields in the associated data, the Equasis ship database and Clarksons ship information platform were used to query and complete them based on the IMO number, ensuring a ship dimension field completeness rate of ≥98%. Missing sea ice thickness data was labeled "0cm" and supplemented with the "ice-free environment" attribute tag. When records from different data sources for the same accident contradicted each other, the Lloyd's Register database record was used as the standard, combined with cross-validation using accident investigation reports, to correct the accident type labeling. The merged dataset was then converted into a structured CSV format to provide standardized data support for subsequent modeling.

[0041] Based on the chi-square test, candidate influencing factors that are significantly related to the consequences of the accident were screened. Multicollinearity among the candidate influencing factors was eliminated by the variance inflation factor test, and the final explanatory variable system was determined. The consequences of the accident include serious accidents and pollution accidents.

[0042] The variable selection process is as follows: Candidate factors are defined as ship factors (flag state, ship type, gross tonnage and ship age), accident factors (accident type and accident year) and environmental factors (sea ice concentration, sea ice thickness and wind speed). Serious accidents are defined as ≥1 death or property damage ≥1 million USD or total loss of ship, and pollution accidents are defined as oil spills ≥10 tons or the spread of hazardous substances.

[0043] SPSS software was used to calculate the chi-square statistics and probability values ​​of each candidate influencing factor and serious accidents and pollution accidents. Candidate influencing factors with probability values ​​< 0.1 were screened to proceed to the next step, while non-significant factors with probability values ​​≥ 0.1 were eliminated. Stata software was used to calculate the variance inflation factor (VIF) of the screened candidate influencing factors. For example, the VIF of sea ice concentration and sea ice thickness was 3.2, indicating no serious collinearity. Factors with VIF ≥ 10 were eliminated, and finally, an explanatory variable system containing multiple influencing factors was formed.

[0044] The specific steps for determining the key influencing factors corresponding to the risk scenario at each spatiotemporal point using a bivariate Probit model include: Maximum likelihood estimation is performed on bivariate Probit models under different risk scenarios, outputting the coefficients, standard deviations, Z-statistics, and probability values ​​of each explanatory variable. Significant variables under each risk scenario are then identified, revealing the influencing factors. The bivariate Probit model is as follows: In the formula, As a latent variable for the severity of serious accidents, To and The corresponding set of explanatory variables, As a latent variable representing the degree of pollution in a pollution incident, To and The corresponding set of explanatory variables, For the regression coefficient vector, For the disturbance term, It is a joint standard normal distribution. For disturbance terms The mean, For disturbance terms The mean, For disturbance terms variance For disturbance terms variance The correlation coefficient between serious accidents and pollution accidents; Serious accidents include: The pollution incident was as follows: In the formula, This is a serious accident. For pollution incident; Calculate the marginal effect of each influencing factor on the probability of simultaneous occurrence of severe accidents and pollution accidents, and rank the influencing factors according to the absolute value of their marginal effects to identify key influencing factors. The formula for calculating the marginal effect of each influencing factor on the probability of simultaneous occurrence of severe accidents and pollution accidents is as follows: In the formula, For the first One influencing factor The marginal effect of the probability of a serious accident and a pollution accident occurring simultaneously. For the set of influencing factors, In the set of influencing factors The probability of serious accidents and pollution accidents occurring simultaneously under certain conditions. Let be the probability density function of a bivariate normal distribution. For the first One influencing factor The direct impact coefficient on serious accidents. For the first One influencing factor The direct impact coefficient on pollution accidents.

[0045] S4. Based on static attribute information and navigation plan information, obtain the navigation behavior characteristics of the target vessel on its planned voyage.

[0046] Navigation behavior characteristics include accident type tendency characteristics, operational characteristics at key points along the route, and spatiotemporal exposure characteristics. Accident type tendency characteristics refer to the historical probability distribution of various types of accidents occurring in the same risk scenarios for vessels with the same static attribute information as the target vessel.

[0047] The key operational characteristics of a route are the pre-defined navigation operations required for a target vessel to handle high-risk sections of its planned route. High-risk sections include narrow waterways, fixed ice perimeters, and areas known to be prone to extreme weather.

[0048] The spatiotemporal exposure characteristics are the exposure duration and path of the target vessel in different risk scenarios. The exposure duration is the cumulative sailing time of the target vessel in each risk scenario. The exposure path is the length of the path taken by the target vessel through different risk scenarios.

[0049] S5. Using the navigation behavior characteristics of the target vessel on its planned voyage and the key influencing factors corresponding to the risk scenario at each spatiotemporal point as input, the joint probability of the target vessel experiencing different types of accident consequences at each spatiotemporal point is obtained through the sub-scenario accident causation model.

[0050] The joint probability of different types of accident consequences occurring at various points in time and space is: In the formula, The joint probability of different types of accident consequences occurring at various spatiotemporal points. This is a serious accident. For pollution incidents, For the first One risk scenario, As a key influencing factor, For navigation behavior characteristics, For normalization function, For risk scenarios The weight parameters of the output layer, For activation function, For risk scenarios The weight parameters of the hidden layer, For splicing operations, For risk scenarios The bias parameters of the hidden layer, For risk scenarios The bias parameters of the output layer.

[0051] By splicing, key influencing factors and navigation behavior characteristics are merged into a comprehensive feature vector, which fully covers the static attributes, environmental conditions, and dynamic behavioral characteristics that affect the consequences of an accident.

[0052] The joint probabilities of different types of accident consequences occurring at various spatiotemporal points include the first joint probability that the target vessel has a serious accident and a pollution accident at the spatiotemporal point, the second joint probability that it has a serious accident but no pollution accident, and the third joint probability that it has no serious accident but a pollution accident.

[0053] Finally, the output contains a list of the three joint probability values ​​for each spatiotemporal point. This provides a direct probabilistic basis for subsequent risk value quantification and level classification.

[0054] S6. Based on the joint probability of different types of accident consequences occurring at various time and space points of the target vessel, quantify the risk value of the target vessel at various time and space points in the planned voyage, and classify the risk level according to the preset threshold to obtain the accident risk assessment result.

[0055] The risk values ​​of the target vessel at various points in time and space during the planned voyage are: In the formula, For the target vessel during the planned voyage Risk value at each spatiotemporal point For the target vessel during the planned voyage The first joint probability of each spatiotemporal point The first joint probability weight, For the target vessel during the planned voyage The second joint probability of each spatiotemporal point For the second joint probability weight, For the target vessel during the planned voyage The third joint probability of each spatiotemporal point The third joint probability weight; The accident risk assessment results are as follows: In the formula, The low-risk threshold The threshold for medium risk. This is a high-risk threshold.

[0056] When the accident risk assessment result is low risk, it means that the risk is within an acceptable range and the ship can sail as originally planned. This usually corresponds to normal navigation status. The crew can maintain routine lookout and operation without taking any additional risk mitigation measures.

[0057] When the accident risk assessment result is medium risk, it indicates that the risk has increased. It is recommended that the crew be more vigilant and prepare emergency plans. This usually corresponds to strengthening the state of alert. The crew should be more vigilant, check and familiarize themselves with the relevant emergency plans, and consider possible preventive measures, such as fine-tuning the course or speed.

[0058] When the accident risk assessment result is high risk, it means that the risk is significant and the ship is facing a clear and high danger. Proactive measures need to be taken to intervene and trigger a risk warning. The captain needs to assess the situation and implement the predetermined risk mitigation plan, such as significantly reducing speed and changing the planned route to avoid high-risk areas.

[0059] When the accident risk assessment result is extremely high risk, it means that the ship is in an extremely dangerous situation, the probability of an accident is high and the consequences are extremely serious, triggering an emergency alarm. The ship should avoid entering or immediately leave the area. If the ship is already in the area, it should take all necessary measures to ensure the safety of the ship, including stopping navigation and seeking external assistance.

[0060] Example 2 This embodiment provides a ship accident risk assessment system suitable for Arctic waters, such as... Figure 2 As shown, it includes: The information acquisition module is used to acquire static attribute information and navigation plan information of the target vessel; The spatiotemporal segmentation module is used to divide the planned voyage of the target vessel into multiple spatiotemporal points based on the voyage plan information, and to determine the risk scenario to which each spatiotemporal point belongs. The key influencing factor identification module is used to determine the key influencing factors corresponding to the risk scenario to which each spatiotemporal point belongs by using a bivariate Probit model based on the environmental data of each spatiotemporal point. The navigation behavior feature acquisition module is used to obtain the navigation behavior features of the target vessel on its planned voyage based on static attribute information and navigation plan information. The joint probability acquisition module is used to take the key influencing factors and navigation behavior characteristics corresponding to the risk scenario of each spatiotemporal point as input, and obtain the joint probability of the target ship experiencing different types of accident consequences at each spatiotemporal point through the sub-scenario accident causation model. The accident risk assessment module is used to quantify the risk value of the target vessel at each time and space point in the planned voyage based on the joint probability of different types of accident consequences occurring at each time and space point, and to classify the risk level according to the preset threshold to obtain the accident risk assessment result.

[0061] The rest is the same as in Example 1.

[0062] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A method for assessing ship accident risks applicable to Arctic waters, characterized in that, Includes the following steps: Obtain the static attribute information and navigation plan information of the target vessel; Based on the navigation plan information, the target vessel's planned voyage is divided into multiple spatiotemporal points, and the risk scenario associated with each spatiotemporal point is determined. Based on the environmental data at each spatiotemporal point, the key influencing factors corresponding to the risk scenario to which it belongs are determined by a bivariate Probit model. Based on the static attribute information and the navigation plan information, the navigation behavior characteristics of the target vessel on its planned voyage are obtained; The navigation behavior characteristics of the target vessel on its planned voyage and the key influencing factors corresponding to the risk scenario at each time and space point are used as inputs. The joint probability of the target vessel experiencing different types of accident consequences at each time and space point is obtained through the sub-scenario accident causation model. Based on the combined probability of different types of accident consequences occurring at various time and space points for the target vessel, the risk value of the target vessel at various time and space points in the planned voyage is quantified, and the risk level is classified according to the preset threshold to obtain the accident risk assessment result.

2. The method for ship accident risk assessment applicable to Arctic waters according to claim 1, characterized in that, The static attribute information includes the vessel's IMO number, flag state, vessel type, gross tonnage, and age. The voyage plan information includes the planned route, planned speed, and start and end times of the voyage.

3. The method for ship accident risk assessment applicable to Arctic waters according to claim 1, characterized in that, The risk scenarios are divided according to seasonal and latitudinal dimensions, including summer-high latitude scenarios, summer-low latitude scenarios, non-summer-high latitude scenarios, and non-summer-low latitude scenarios.

4. The method for ship accident risk assessment applicable to Arctic waters according to claim 1, characterized in that, The environmental data includes sea ice data and meteorological data. The sea ice data includes sea ice concentration and sea ice thickness, and the meteorological data includes wind speed data at a height of 10 meters above the sea surface.

5. The method for assessing ship accident risks in Arctic waters according to claim 1, characterized in that, The bivariate Probit model uses severe accidents and pollution accidents as dependent variables and the final explanatory variable system as independent variables. The specific steps for determining the key influencing factors corresponding to the risk scenario at each spatiotemporal point using the bivariate Probit model include: Maximum likelihood estimation is performed on bivariate Probit models under different risk scenarios, outputting the coefficients, standard deviations, Z-statistics, and probability values ​​of each explanatory variable. Significant variables under each risk scenario are then identified, revealing the influencing factors. The bivariate Probit model is as follows: In the formula, As a latent variable for the severity of serious accidents, To and The corresponding set of explanatory variables, As a latent variable representing the degree of pollution in a pollution incident, To and The corresponding set of explanatory variables, For the regression coefficient vector, For the disturbance term, It is a joint standard normal distribution. For disturbance terms The mean, For disturbance terms The mean, For disturbance terms variance For disturbance terms variance The correlation coefficient between serious accidents and pollution accidents; The serious accident is as follows: The pollution incident was as follows: In the formula, This is a serious accident. For pollution incident; Calculate the marginal effect of each influencing factor on the probability of the simultaneous occurrence of serious accidents and pollution accidents, and sort the influencing factors according to the magnitude of the absolute value of the marginal effect to screen out the key influencing factors.

6. The method for assessing ship accident risks in Arctic waters according to claim 5, characterized in that, The formula for calculating the marginal effect of each influencing factor on the probability of the simultaneous occurrence of serious accidents and pollution accidents is as follows: In the formula, For the first One influencing factor The marginal effect of the probability of a serious accident and a pollution accident occurring simultaneously. For the set of influencing factors, In the set of influencing factors The probability of serious accidents and pollution accidents occurring simultaneously under certain conditions. Let be the probability density function of a bivariate normal distribution. For the first One influencing factor The direct impact coefficient on serious accidents. For the first One influencing factor The direct impact coefficient on pollution accidents.

7. The method for assessing ship accident risks in Arctic waters according to claim 1, characterized in that, The joint probability of different types of accident consequences occurring at various points in time and space is: In the formula, The joint probability of different types of accident consequences occurring at various spatiotemporal points. This is a serious accident. For pollution incidents, For the first One risk scenario, As a key influencing factor, For navigation behavior characteristics, For normalization function, For risk scenarios The weight parameters of the output layer, For activation function, For risk scenarios The weight parameters of the hidden layer, For splicing operations, For risk scenarios The bias parameters of the hidden layer, For risk scenarios The bias parameters of the output layer.

8. The method for ship accident risk assessment applicable to Arctic waters according to claim 7, characterized in that, The joint probabilities of different types of accident consequences occurring at various spatiotemporal points include the first joint probability that the target vessel has a serious accident and a pollution accident at the spatiotemporal point, the second joint probability that it has a serious accident but no pollution accident, and the third joint probability that it has no serious accident but a pollution accident.

9. The method for ship accident risk assessment applicable to Arctic waters according to claim 8, characterized in that, The risk values ​​of the target vessel at various points in time and space during its planned voyage are: In the formula, For the target vessel during the planned voyage Risk value at each spatiotemporal point For the target vessel during the planned voyage The first joint probability of each spatiotemporal point The first joint probability weight, For the target vessel during the planned voyage The second joint probability of each spatiotemporal point For the second joint probability weight, For the target vessel during the planned voyage The third joint probability of each spatiotemporal point The third joint probability weight; The accident risk assessment results are as follows: In the formula, The low-risk threshold The threshold for medium risk. This is a high-risk threshold.

10. A ship accident risk assessment system applicable to Arctic waters, characterized in that, include: The information acquisition module is used to acquire static attribute information and navigation plan information of the target vessel; The spatiotemporal division module is used to divide the planned voyage of the target vessel into multiple spatiotemporal points based on the navigation plan information, and to determine the risk scenario to which each spatiotemporal point belongs. The key influencing factor identification module is used to determine the key influencing factors corresponding to the risk scenario to which each spatiotemporal point belongs by using a bivariate Probit model based on the environmental data of each spatiotemporal point. The navigation behavior feature acquisition module is used to obtain the navigation behavior features of the target vessel on its planned voyage based on the static attribute information and the navigation plan information. The joint probability acquisition module is used to take the key influencing factors and navigation behavior characteristics corresponding to the risk scenario of each spatiotemporal point as input, and obtain the joint probability of the target ship experiencing different types of accident consequences at each spatiotemporal point through the sub-scenario accident causation model. The accident risk assessment module is used to quantify the risk value of the target vessel at each time and space point in the planned voyage based on the joint probability of different types of accident consequences occurring at each time and space point, and to classify the risk level according to the preset threshold to obtain the accident risk assessment result.

Images