Ship navigation risk control method

Abstract

The invention belongs to the technical field of ship navigation risk management, and particularly relates to a ship navigation risk control method. Comprising the following steps: acquiring a ship prepared standard data file through a ship navigation service system; determining corresponding elements under the accident type; sorting the comprehensive weight and the influence weight to form a risk accident element graph model; extracting historical navigation accident data according to element types in the risk accident element graph model; establishing a graph form sample database; a multi-layer network classification model is adopted to carry out classification training on the navigation map form samples; and matching ship risk accident features to be analyzed to carry out risk management and control. The method is used for completing risk accident feature classification and identification of different types of ships in the sailing process by analyzing various element features in the historical data of the ship sailing risk accidents, performing classification prediction on the to-be-analyzed ship sailing risk accidents, and determining the sailing accident features so as to take targeted accident prevention measures.

Description

Technical Field

[0001] The present invention belongs to the technical field of ship navigation risk management, and in particular relates to a ship navigation risk control method. Background Art

[0002] Ship navigation accidents often lead to significant casualties and property losses. The analysis and control of ship navigation accidents often involve a large number of influencing factors, including but not limited to the ship's own status, weather conditions, channel conditions, etc. The occurrence of each navigation accident often involves the combined influence of multiple factors. The current risk control of ship navigation accidents still mainly relies on ship operators, who conduct analysis and judgment based on past historical experience combined with the current ship status and route status, supplemented by various standardized risk control processes. Summary of the invention

[0003] The object of the present invention is to provide a method for controlling ship navigation risks for realizing intelligent and automated ship navigation risk analysis so as to determine the focus of ship risk control and select a suitable control scheme.

[0004] To achieve the above purpose, the present invention adopts the following technical solution.

[0005] A method for controlling ship navigation risks comprises the following steps:

[0006] Step 1: Obtain navigation accident data information during the navigation process of ships through the ship navigation service system, formulate standard data files according to the needs of navigation risk influencing factors, and summarize the attribute data in various ship navigation service systems through standard data files;

[0007] Step 2: Determine the corresponding elements under the categories of improper operation accidents, ship problem accidents, and route problem accidents; specifically:

[0008] Extract historical data of navigation accidents caused by improper ship operation, and sort out the element sequence {X1,X2,X3,...X i ,...X I}, where X i refers to the i-th element, and I refers to the total number of samples of the element;

[0009] For a certain type of improper operation accident Y j , analyze each improper operation accident Y j Causes of accidents caused by improper operation Y j The sequence of elements that occur {X jk1 ,X jk2 ,...X jki ,...X jkI};

[0010] Where X jki Refers to the kth improper operation accident Y j When the element X occurs i The indicated value, X jki =1 indicates the kth improper operation accident Y j When the occurrence occurs, there is factor X i The impact of X jki =0 indicates the kth improper operation accident Y j Occurs when feature X does not exist i The impact of

[0011] Get all the element sequences and delete X jki = 0, and then remove the repeated element sequence to obtain the accident Y that caused the improper operation j The element combination sequence T j ={T j1 ,T j2 ...T jm ...T jM}; where T jm Indicates an accident caused by improper operation Y j The mth element set,

[0012] Extract risk accident sample set B affected by any combination of factors from historical data j , based on expert scoring system or statistical analysis, statistical T jm The combination of medium factors for risk accident sample set B j The number of samples whose influence degree is the lowest among all factor combinations is b jm.min , calculate T jm The influence coefficient under the corresponding factor combination where b j Refers to sample set B j The total number of samples in the sample; calculate the improper operation accident Y j The comprehensive weight of the combination of factors involved

[0013] Based on the above steps, calculate the ship problem accident O p Combination of ship problem elements O pq Corresponding factor combination influence coefficient and element combination sequence O p = {O p1 ,O p2 ... pq ... pQ The comprehensive weight of Calculate route problem accident U v The combination of route problem elements U vsThe corresponding factor combination influence coefficient and the factor combination sequence U v ={U v1 ,U v2 ...U vs ...U vS} of the comprehensive weight

[0014] where O pq refers to the qth ship problem element combination corresponding to the pth type of ship problem accident, and b p refers to the total number of accident samples B p in the historical data affected by any combination of ship problem elements, and b pq.min is the number of samples in the sample set B p where the influence degree of the ship problem element combination O pq is the lowest among all ship problem element combinations; where U vs refers to the sth route problem element combination corresponding to the vth type of route problem accident, and b v refers to the total number of accident samples B v in the historical data affected by any combination of route problem elements, and b vs.min is the number of samples in the sample set B v where the influence degree of the route problem element combination U vs is the lowest among all route problem element combinations;

[0015] Step 3: Sort the comprehensive weight and the influence weight respectively. For each type of ship, sort the improper operation accidents, ship problem accidents, and route problem accidents in descending order of weight. At the same time, screen and sort the corresponding combined elements within the improper operation accidents, ship problem accidents, and route problem accidents, and respectively screen several element combinations with larger influence weights from each type of accident in descending order to form the risk accident element diagrams of various types of ships;

[0016] Step 4: Extract the historical navigation accident data of the corresponding ships according to the element types in the risk accident element diagrams of various types of ships; for each ship navigation accident sample, extract the element index value y i ′ and normalize it to obtain the normalized value y i ;

[0017]

[0018] where y i ′ .max is the maximum value of the element index y i ′, and y i ′ .min is the element index y iThe minimum value of ′

[0019] According to the order of the risk accident element diagrams of the corresponding ship types, after sorting, the element sequence Y of each ship is obtained w ={y w1 , y w2 , y w3 ,... y wl ,... y wL}; where y wl refers to the index value of the l-th element in the risk accident element diagram of the w-th ship, and L refers to the total number of elements in the risk accident element diagram

[0020] Convert the element sequence Y w uniformly into a graphical sample Sort all the graphical samples in each type of ship to establish a navigation graphical sample database for this type of ship where H×W is the size of the graph

[0021] Step 5: Use a multi-layer network classification model to classify and train the navigation graphical samples; the multi-layer network classification model is based on the ResNet34 multi-branch network and includes a feature extraction module and a feature fusion module

[0022] The input of the feature extraction module is the graphical sample Through the multi-layer dilated convolution blocks of the ResNet34 network, output images with different dilation rates are generated, and the features of the output images are extracted through the Inception network

[0023] The feature fusion module uses a multi-pole feature weighted fusion network for feature weighted fusion, including outputting fusion features after dilated convolution, normalization, rectified linear activation, and adaptive average pooling of the feature maps of the output images with different dilation rates

[0024] Divide the graphical sample database into a training data set and a validation data set, use the training data set to train the multi-layer network classification model, and use the validation data set to optimize and adjust the parameters of the multi-layer network classification model to obtain the final multi-layer network classification model

[0025] Step 6: Input the samples in the graphical sample database into the multi-layer network classification model that has completed training to obtain classification results of several types, obtain the typical feature categories representing the navigation accidents of a certain type of ship, and the navigation accident features of ships with similar basic element features to the classification results are consistent with this typical feature category

[0026] By collecting the information on the past navigation accident elements of the ship to be analyzed and inputting it into the classification network for classification, the typical characteristics of its navigation accidents can be obtained. By matching the risk accident characteristics of similar ships with the same typical characteristics on the expected route of the ship to be analyzed, the navigation risk accidents that may occur to the ship to be analyzed can be predicted and risk control can be carried out according to the prediction results.

[0027] For further improvement or preferred implementation steps of the aforementioned ship navigation risk control method, the ship navigation service system includes various ship navigation service systems such as VTS system, GPS-CDMA, and AIS.

[0028] For further improvement or preferred implementation steps of the aforementioned ship navigation risk control method, the standard data file includes data blocks with the following data indicators:

[0029] Code index data block: Used to save the code serial number data of each ship as the unique identification code of each ship;

[0030] Type index data block: Used to save the type information of each ship. The ship types at least include small ships, medium-sized ships, large ships, unmanned ships, and special ships. Among them, small ships refer to ships with an overall length less than or equal to 20m, medium-sized ships refer to ships with an overall length greater than 20m and less than or equal to 100m, large ships refer to ships with an overall length exceeding 100m, unmanned ships refer to ships that mainly complete navigation based on autonomous driving technology, and special ships refer to ships for specific purposes such as engineering ships, scientific research ships, and rescue ships;

[0031] Speed index data block: Used to save the speed information of each ship during navigation;

[0032] Course index data block: Used to save the route data information of each ship during navigation;

[0033] Lifecycle index data block: Used to save the average lifecycle of each ship and the information on its lifecycle stage;

[0034] Water depth index data block: Used to save the water depth data information of the ship at each navigation node or section;

[0035] Resistance index data block: Used to save the ship navigation resistance index, which refers to the ratio of the resistance generated by the water surface wind force and the underwater fluid force to the ship's current course during the navigation of the ship at each navigation section or node to the ship's power;

[0036] Visibility index data block: Used to save the visibility range data of the ship at each navigation node or section;

[0037] Unmanned ship index data block: used to store the number of unmanned ships on each current route and their proportion data;

[0038] Route obstacle index data block: used to store the obstacle distribution data on each section of the route;

[0039] Risk ship index data block: used to store the proportion data of high-risk ships on each route, where the high-risk ships refer to those that will cause serious consequences in case of dangerous situations, or the proportion of ships with high navigation risk in the current section of the route.

[0040] Further improvement or preferred implementation steps of the aforementioned ship navigation risk control method, and the standard data file is saved through a unified CSV text file.

[0041] Further improvement or preferred implementation steps of the aforementioned ship navigation risk control method, and step 1 further includes cleaning and complementing the original data. Specifically:

[0042] A1. Since ship navigation data is continuous in time series, for single-point missing data, interpolation and averaging are performed using the data at the moments before and after the sampling point for filling. For continuous missing data, data fitting within the sampling period or averaging fitting with historical sampling data of the same period is used to obtain the data;

[0043] A2. For the original data after filling, it is necessary to analyze and judge its quality to ensure that the provided data has sufficient accuracy. Specifically, it includes:

[0044] Verify the consistency of all data, verify whether the data has the required type, order of magnitude, format, and unit, and check and analyze inconsistent data, and mark the error data; including: whether numerical indicators such as speed have a relatively consistent order of magnitude; whether code-type data such as dates, models, and serial numbers have a consistent text format;

[0045] Perform fitting analysis on multi-sample data of a single indicator to judge whether the indicator data exceeds the normal and reasonable range, and mark the data that exceeds the normal and reasonable range as an error, including: numerical data exceeding the normal upper limit; code-type data codes exceeding the normal code range, etc.;

[0046] A3. For the preprocessed raw data, we need to collect the complete data of a single ship during each voyage. Since the ship is recorded in different voyage service systems, a large amount of duplicate data will exist during the initial data collection process. These duplicate data show random fluctuations within a certain range at their respective nodes. This kind of fluctuation is caused by differences in monitoring methods, monitoring time nodes, data storage accuracy, etc. of various voyage service systems. Only one set of each group of duplicate data is needed as the raw data for analysis. In order to retain the basic data characteristics of these data, the data is compressed through a clustering analysis algorithm. Brief Description of the Drawings

[0047] Figure 1 It is a schematic diagram of the multi-layer network classification model structure. Detailed Implementation Manner

[0048] The following describes the present invention in detail with specific embodiments.

[0049] The ship navigation risk control method of the present application is mainly used to provide a method that can complete the classification and identification of risk accident characteristics of different types of ships during navigation by analyzing the characteristics of various elements in the historical data of ship navigation risk accidents, and thereby classify and predict the ship navigation risk accidents to be analyzed, determine their navigation accident characteristics, so as to take targeted accident prevention measures.

[0050] It mainly includes the following steps:

[0051] Step 1: Obtain basic navigation elements

[0052] The current ship navigation service system has established a perfect ship navigation information and data management process by combining means such as sound, light, electric communication measures and radar communication. Through various ship navigation service systems such as VTS systems, GPS-CDMA, and AIS, it is possible to obtain the navigation accident data information during the ship's navigation. As the control requirements for navigation risks to be considered in this application, it is necessary to extract the corresponding element information from the data information provided by the above ship navigation service systems for subsequent use and processing;

[0053] Since various ship navigation service systems usually have their own independent file data formats, for the convenience of unified processing, a standard data file should first be drawn up according to the requirements of navigation risk influencing factors, and the attribute data in various ship navigation service systems is summarized through the standard data file as the raw data required in this application for extracting various basic navigation elements;

[0054] In particular, to meet the most basic requirements, the standard data file in this application should at least include data blocks with the following data indicators:

[0055] Code Index Data Block: Used to store the code serial number data of each ship as the unique identification code for each ship;

[0056] Type Index Data Block: Used to store the type information of each ship. The ship types at least include small ships, medium ships, large ships, unmanned ships, and special ships. Among them, small ships refer to ships with an overall length less than or equal to 20m, medium ships refer to ships with an overall length greater than 20m and less than or equal to 100m, large ships refer to ships with an overall length exceeding 100m, unmanned ships refer to ships that mainly complete navigation based on autonomous driving technology, and special ships refer to ships for specific purposes such as engineering ships, scientific research ships, and rescue ships;

[0057] Speed Index Data Block: Used to store the speed information of each ship during navigation;

[0058] Course Index Data Block: Used to store the route data information of each ship during navigation;

[0059] Lifecycle Index Data Block: Used to store the average lifecycle of each ship and the information of its lifecycle stages;

[0060] Water Depth Index Data Block: Used to store the water depth data information of the ship at each navigation node or section;

[0061] Resistance Index Data Block: Used to store the ship navigation resistance index, and the resistance index refers to the ratio of the resistance generated by the water surface wind force and the underwater fluid force to the ship's current course during the navigation of each section or node of the ship to the ship's power;

[0062] Visibility Index Data Block: Used to store the visibility range data of the ship at each navigation node or section;

[0063] Unmanned Ship Index Data Block: Used to store the number and proportion data of unmanned ships on each current route;

[0064] Route Obstacle Index Data Block: Used to store the obstacle distribution data on each section of the route;

[0065] Risk Ship Index Data Block: Used to store the proportion data of high-risk ships on each route. High-risk ships refer to ships that will cause serious consequences in case of dangerous situations, or the proportion of ships with high navigability risks in the current section of the ship;

[0066] The basic data of various ship navigation service systems are usually stored in the form of various text data tables. During the above-mentioned process of processing and integrating the original data, it is considered to convert and save them as a unified CSV text file for storage, so as to facilitate the reading and processing by various data analysis and processing software, simplify the processing difficulty of the original data, and improve the efficiency of data screening and analysis;

[0067] A large number of common data are saved and processed in the basic data of various ship navigation service systems. Due to factors such as data accuracy, recording time, and deviation of detection positions, there may be differences in the data of the same index. At the same time, due to various reasons, there are also cases where the data of specific nodes or moments are missing; Therefore, this application also includes the cleaning and complementing process of the original data. Specifically:

[0068] A1. Since ship navigation data is continuous in time series, for single-point missing data, it can be filled by taking the difference and averaging the moments before and after the sampling point. For continuous missing data, it is obtained by fitting the data within the sampling period or averaging and fitting the historical sampling data of the same period;

[0069] A2. For the original data after filling, it is necessary to analyze and judge its quality to ensure that the provided data has sufficient accuracy. Specifically, it includes:

[0070] Verify the consistency of all data, verify whether the data has the required type, order of magnitude, format, and unit, and check and analyze the inconsistent data and mark the error data; including: whether the numerical indicators such as speed have a relatively consistent order of magnitude; whether the coded data such as dates, models, and serial numbers have a consistent text format;

[0071] Perform fitting analysis on the multi-sample data of a single index to judge whether the index data exceeds the normal and reasonable range, and mark the data that exceeds the normal and reasonable range as errors, including: numerical data exceeding the normal upper limit; coded data codes exceeding the normal code range, etc.;

[0072] A3. For the preprocessed original data, it is necessary to collect the complete data of a single ship during each voyage. Since the ship may have records in different navigation service systems, a large number of duplicate data will be generated during the initial data collection process. These duplicate data show random fluctuations within a certain range at their respective nodes. This kind of fluctuation is caused by differences in monitoring methods, monitoring time nodes, data storage accuracy, etc. of various navigation service systems. Only one set of each group of duplicate data is required as the original data for analysis. To retain the basic data characteristics of these data, the data is compressed through a clustering analysis algorithm.

[0073] Although ship navigation accidents are low-probability events in the navigation database, they are diverse in types, including meeting-ship collision accidents, obstacle collision accidents, grounding and capsizing accidents, fire and explosion accidents, wind disasters, self-sinking accidents, flooding accidents, other accidents, etc. Analyzing the causes of various accidents can be attributed to accidents caused by improper ship operation and accidents caused by problems with the ship or route; based on the accident tree analysis principle, it can be determined that the accident tree with ship navigation accidents as the final event can be divided into improper operation accidents, ship problem accidents, and route problem accidents;

[0074] Each basic navigation element will cause various risks of ship navigation and lead to the occurrence of accidents as basic elements. For example, accidents caused by improper ship operation are mainly affected by some improper or incorrect operation behaviors during the ship operation process. By analyzing the characteristics of various faults and the correlation between ship operation behaviors, the corresponding elements of improper ship operation accidents can be sorted out and analyzed. The common elements mainly include: ship speed, ship acceleration, ship heading, variable angular velocity, draft depth, etc. The meanings of each element are shown in Table 1.

[0075] Table 1 Meanings of elements and improper operations

[0076]

[0077] Taking the 62 ship accidents caused by improper operation that occurred to a certain type of ship in a certain navigation section in the past three years as an example, the operation data was extracted based on the aforementioned elements respectively to obtain Table 2. At the same time, 62 samples were randomly selected from the same type of ships passing normally, and their operation data was extracted to obtain Table 3.

[0078] Table 2 Element data corresponding to improper operations

[0079]

[0080] Table 3 Normal element data

[0081]

[0082] Through comparative analysis, it can be seen that compared with the element data of normal operation, the change interval ranges of ship speed, acceleration, and heading deviation in the elements corresponding to improper operation accidents are significantly larger. From the variance and standard deviation data, the fluctuations of the aforementioned elements also exceed normal operation, indicating that during the occurrence of improper operation accidents, the navigation speed and operation volume are larger. Compared with normal operation, the change intervals and fluctuations of ship heading deviation, variable angular velocity, etc. in the elements corresponding to improper operation accidents are also relatively high, indicating that during the occurrence of improper operation accidents, the operator's control of the corresponding elements is relatively radical, with a larger operation volume and more operation times. The above operation characteristics are directly related to the operation habits of ship operators, etc.

[0083] Further statistical analysis reveals that for a group of operators with similar operating habits, during the operation of a ship, the probability of similar or analogous navigation accidents occurring at the same location and in the same state is also closer. Through various elements extracted from the original data, the navigation risks of operators of the same type when operating the corresponding ships to perform similar tasks can be evaluated, and thus risk planning can be carried out in advance.

[0084] Generally speaking, a single inappropriate element does not necessarily lead to the occurrence of a risk accident. Usually, multiple inappropriate operations occur simultaneously or continuously to cause a risk accident, and at the same time, the superposition of multiple elements leads to the occurrence of an improper operation accident. To analyze the possibility of an accident occurring, it is necessary to determine the importance of each element for the risk accident.

[0085] Extract the historical data of navigation accidents caused by improper ship operations, and organize to obtain the element sequence {X1, X2, X3,... X i ,... X I} corresponding to the improper operation accident, where X i refers to the i-th element, and I refers to the total sample number of elements;

[0086] For a certain type of improper operation accident Y j , analyze the cause of each improper operation accident Y j , and extract the element sequence {X j , X jk1 ,... X jk2 ,... X jki ,... X jkI} that causes the occurrence of the improper operation accident Y

[0087] Among them, X jki refers to the indication value of element X j when the k-th improper operation accident Y i occurs. X jki = 1 indicates the existence of the influence of element X j when the k-th improper operation accident Y i occurs, and X jki = 0 indicates the non-existence of the influence of element X j when the k-th improper operation accident Y i occurs;

[0088] Obtain all the element sequences, delete the elements with X jki = 0, and then eliminate the duplicate element sequences to obtain the element combination sequence T j = {T j , T j1 ,... T j2 ,... T jm ,... T jM}; where T jm represents the m-th set of elements that cause the improper operation accident Y j of,

[0089] Extract the risk accident sample set B affected by any combination of elements from historical data j , and based on the expert scoring system or statistical analysis, count T jm the number b of samples in the risk accident sample set B j whose degree of influence by the combination of elements in is the lowest among all combinations of elements jm.min , calculate the influence coefficient corresponding to the combination of elements of T jm where b is the total number of samples in the sample set B j ; calculate the comprehensive weight of the combination of elements involved in the improper operation accident Y j where b j is the total number of samples in the sample set B

[0090] Based on the above steps, calculate the ship problem accident O p the combination of ship problem elements O pq corresponding to the influence coefficient of the combination of elements and the comprehensive weight of the combination of element sequences O p ={O p1 , O p2 ... O pq ... O pQ} Calculate the route problem accident U v the combination of route problem elements U vs corresponding to the influence coefficient of the combination of elements and the comprehensive weight of the combination of element sequences U v ={U v1 , U v2 ... U vs ... U vS}

[0091] where O pq refers to the q-th ship problem element combination corresponding to the p-th type of ship problem accident, b p refers to the total number of accident samples B in historical data affected by any combination of ship problem elements p ; b pq.min is the sample set B p the number of samples in which the influence degree of the ship problem element combination O pq is the lowest among all ship problem element combinations; where U vs refers to the s-th route problem element combination corresponding to the v-th type of route problem accident, bv refers to the accident sample set B affected by any combination of elements in the route problem in historical data v the total number of samples, b vs.min is the sample set B v in which the combination of route problem elements U vs the degree of influence belongs to the number of samples with the lowest among all combinations of route problem elements;

[0092] Based on the foregoing steps, determine the comprehensive weights of improper operation accidents, ship problem accidents, and route problem accidents corresponding to different types of ships, as well as the influence weights of the combination of elements corresponding to improper operation accidents, ship problem accidents, and route problem accidents;

[0093] Sort the comprehensive weights and influence weights respectively. For each type of ship, sort the improper operation accidents, ship problem accidents, and route problem accidents in descending order of weight. At the same time, screen and sort the corresponding combined elements within the improper operation accidents, ship problem accidents, and route problem accidents, and respectively screen several combined elements with larger influence weights from each type of accident in descending order to form the risk accident element diagram model of each type of ship;

[0094] According to the element types in the risk accident element diagram model of each type of ship, extract the historical navigation accident data of the corresponding ship; for each ship navigation accident sample, extract the element index value y i ′ and normalize it to obtain the normalized value y i ;

[0095]

[0096] where y i ′ .max is the maximum value of the element index y i ′, y i ′ .min is the minimum value of the element index y i ′;

[0097] According to the order of the risk accident element diagram model of the corresponding ship type, after sorting, obtain the element sequence Y of each ship w ={y w1 ,y w2 ,y w3 ,...y wl ,...y wL}; where y wl refers to the index value of the l-th element in the risk accident element diagram model of the w-th ship, and L refers to the total number of elements in the risk accident element diagram model;

[0098] Convert the element sequence Y w uniformly into a graphical sample Sort out all graphical samples in each type of ship to establish a navigation graphical sample database for this type of ship. Where H×W is the size of the graph; the graphical samples contain the element characteristics of each ship navigation accident, and all graphical samples of the same type of ship contain the navigability risk accident characteristics of each type of ship. To obtain the response characteristics and establish a navigability risk accident identification scheme for different ships, a multi-layer network classification model is used to classify and train the navigation graphical samples.

[0099] The multi-layer network classification model of this application is as Figure 1 shown. The multi-layer network classification model is based on the ResNet34 multi-branch network and includes a feature extraction module and a feature fusion module;

[0100] The input of the feature extraction module is the graphical sample Through the multi-layer dilated convolution blocks of the ResNet34 network, output images with different dilation rates are generated. The output images are used to extract the features of the output images through the Inception network.

[0101] The feature fusion module uses a multi-pole feature weighted fusion network for feature weighted fusion, including the feature maps of the output images with different dilation rates passing through dilated convolution, normalization, rectified linear activation, and adaptive average pooling to output the fusion features;

[0102] Divide the graphical sample database into a training data set and a validation data set. Use the training data set to train the multi-layer network classification model, and use the validation data set to optimize and adjust the parameters of the multi-layer network classification model to obtain the final multi-layer network classification model;

[0103] Input the samples in the graphical sample database into the trained multi-layer network classification model to obtain classification results of several types, and obtain the typical feature categories representing the navigation accidents of a certain type of ship. The navigation accident characteristics of ships with similar basic element characteristics to the classification results are consistent with this typical feature category;

[0104] By collecting the past navigation accident element information of the ship to be analyzed and inputting it into the classification network for classification, the typical characteristics of its navigation accidents can be obtained. By matching the risk accident characteristics of the same type of ships with the same typical characteristics on the expected route of the ship to be analyzed, the possible navigation risk accidents of the ship to be analyzed can be predicted, and targeted measures can be taken to avoid the occurrence of navigation risk accidents.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting the protection scope of the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the essence and scope of the technical solutions of the present invention.

Claims

1. A method for controlling ship navigation risks, characterized in that: The steps include: Step 1: Obtain navigation accident data information during the navigation process of ships through the ship navigation service system, formulate standard data files according to the needs of navigation risk influencing factors, and summarize the attribute data in various ship navigation service systems through standard data files; Step 2: Determine the corresponding elements under the categories of improper operation accidents, ship problem accidents, and route problem accidents; specifically: Extract historical data of navigation accidents caused by improper ship operation, and sort out the element sequence {X1,X2,X3,...X i ,...X I }, where X i refers to the i-th element, and I refers to the total number of samples of the element; For a certain type of improper operation accident Y j , analyze each improper operation accident Y j Causes of accidents caused by improper operation Y j The sequence of elements that occur {X jk1 ,X jk2 ,...X jki ,...X jkI }; Where X jki Refers to the kth improper operation accident Y j When the element X occurs i The indicated value, X jki =1 indicates the kth improper operation accident Y j When the event occurs, there is element X i The impact of X jki =0 indicates the kth improper operation accident Y j Occurs when feature X does not exist i The impact of Get all the element sequences and delete X jki = 0, and then remove the repeated element sequence to obtain the accident Y that caused the improper operation j The element combination sequence T j ={T j1 ,T j2 ...T jm ...T jM }; where T jm Indicates an accident caused by improper operation Y j The mth element set, Extract risk accident sample set B affected by any combination of factors from historical data j , based on expert scoring system or statistical analysis, statistical T jm The combination of medium factors for risk accident sample set B j The number of samples whose influence degree is the lowest among all factor combinations is b jm.min , calculate T jm The influence coefficient under the corresponding factor combination where b j Refers to sample set B j The total number of samples in the sample; calculate the improper operation accident Y j The comprehensive weight of the combination of factors involved Based on the above steps, calculate the ship problem accident O p Combination of ship problem elements O pq Corresponding factor combination influence coefficient and element combination sequence O p = {O p1 ,O p2 ... pq ... pQ The comprehensive weight of Calculate route problem accident U v The combination of route problem elements U vs Corresponding factor combination influence coefficient And element combination sequence U v = {U v1 ,U v2 ...U vs ...U vS The comprehensive weight of Among them, pq It refers to the qth combination of ship problem elements corresponding to the pth type of ship problem accident, b p It refers to the accident sample set B in the historical data that is affected by any combination of ship problem elements. p The total number of samples, b pq.min is sample set B p Combination of ship problem elements in pq The number of samples whose impact degree is the lowest among all combinations of ship problem factors; vs It refers to the combination of route problem elements in the sth category corresponding to the vth category route problem accident, b v It refers to the accident sample set B in the historical data that is affected by any combination of route problem elements. v The total number of samples, b vs.min is sample set B v Combination of factors in the route problem U vs The number of samples whose impact is the lowest among all combinations of route problem factors; Step 3: Sort the comprehensive weights and impact weights respectively. For each type of ship, sort the improper operation accidents, ship problem accidents, and route problem accidents in descending order of weight. At the same time, screen and sort the corresponding combination elements in the improper operation accidents, ship problem accidents, and route problem accidents. Select several combinations of elements with larger impact weights from each type of accident from large to small to form a risk accident element diagram for each type of ship. Step 4: According to the element types in the risk accident element model of each type of ship, extract the historical navigation accident data of the corresponding ship; for each ship navigation accident sample, extract the element index value y i ′ and normalize to get the normalized value y i ; where y i ' .max is the factor index y i The maximum value of ′, y i ' .min is the factor index y i The minimum value of ′; According to the order of the risk accident factor graph model of the corresponding ship type, the factor sequence Y of each ship is obtained after sorting w ={y w1 ,y w2 ,y w3 ,...y wl ,...y wL }; where y wl It refers to the index value of the lth element in the risk accident element model of the wth ship, and L refers to the total number of elements in the risk accident element model; The element sequence Y w Unified conversion to graph samples Organize all the graphic samples of each type of ship to establish a navigation graphic sample database for that type of ship Where H×W is the size of the graph; Step 5: Use a multi-layer network classification model to classify and train the navigation chart samples; the multi-layer network classification model is established based on the ResNet34 multi-branch network, including a feature extraction module and a feature fusion module; The input of the feature extraction module is a graphical sample The multi-layer dilated convolutional blocks of the ResNet34 network generate output images with different dilation rates, and the output images are passed through the Inception network to extract the features of the output images; The feature fusion module uses a multi-pole feature weighted fusion network to perform feature weighted fusion, including outputting fusion features after dilated convolution, normalization, linear rectification activation, and adaptive average pooling of feature maps of output images with different dilation rates; The graphical sample database The dataset is divided into a training data set and a validation data set. The training data set is used to train the multi-layer network classification model. The validation data set is used to optimize the parameters of the multi-layer network classification model to obtain the final multi-layer network classification model. Step 6: Input the samples in the graphical sample database into the trained multi-layer network classification model to obtain several types of classification results, and obtain a typical feature category representing a navigation accident of a certain type of ship. The navigation accident characteristics of ships with similar basic element characteristics to the classification results are consistent with the typical feature category; By collecting the past navigation accident element information of the ship to be analyzed and inputting it into the classification network for classification, the typical characteristics of its navigation accidents can be obtained. By matching the risk accident characteristics of similar ships with the same typical characteristics on the expected route of the ship to be analyzed, the navigation risk accidents of the ship to be analyzed can be predicted and risk management can be carried out based on the prediction results.

2. The ship navigation risk control method according to claim 1, characterized in that: The ship navigation service system includes various ship navigation service systems such as VTS system, GPS-CDMA, AIS, etc.

3. The ship navigation risk control method according to claim 1, characterized in that: The standard data file includes a data block with the following data indicators: Code index data block: used to store the code serial number data of each ship as the unique identification code of each ship; Type index data block: used to store the type information of each ship, which includes at least small ships, medium-sized ships, large ships, unmanned ships, and special ships; small ships refer to ships with an overall length of less than or equal to 20m, medium-sized ships refer to ships with an overall length of more than 20m and less than or equal to 100m, and large ships refer to ships with an overall length of more than 100m; unmanned ships refer to ships that mainly complete navigation based on autonomous driving technology, and special ships refer to ships with specific purposes such as engineering ships, scientific research ships, and rescue ships; Speed ​​index data block: used to save the speed information of each ship during navigation; Heading index data block: used to save the route data information of each ship; Life cycle indicator data block: used to store the average life cycle of each ship and its life cycle stage information; Water depth index data block: used to save the water depth data information of the ship at each navigation node or section; Resistance index data block: used to store the ship's navigation resistance index, which refers to the ratio of the resistance generated by the surface wind force and underwater fluid force on the ship's current heading during the navigation process of each navigation section or node to the ship's power; Visibility index data block: used to save the visibility range data of the ship at each navigation node or section; Unmanned ship index data block: used to save the number of unmanned ships on each current route and their percentage data; Route obstacle index data block: used to save obstacle distribution data on each route segment; Risky ship index data block: used to save the data on the proportion of high-risk ships on each route. The high-risk ships refer to the proportion of ships that will cause serious consequences when a dangerous situation occurs, or the proportion of ships with high navigation risks in the current section.

4. The ship navigation risk control method according to claim 1, characterized in that: The standard data file is saved via a unified CSV text file.

5. The ship navigation risk control method according to claim 1, characterized in that: The step 1 also includes cleaning and completing the original data, specifically: A1. Since the ship navigation data is continuous in time series, for single-point missing data, it is filled by performing difference and average processing before and after the sampling point. For continuous missing data, it is obtained by fitting the data within the sampling period or averaging the historical sampling data in the same period; A2. The quality of the original data after filling needs to be analyzed and judged to ensure that the data provided has sufficient accuracy, including: Verify the consistency of all data, verify whether the data has the required type, magnitude, format, and unit, and check and analyze inconsistent data, and mark erroneous data; including: whether numerical indicators such as speed have relatively consistent magnitudes; whether code data such as date, model, serial number, etc. have consistent text formats; Perform fitting analysis on multiple sample data of a single indicator to determine whether the indicator data exceeds the normal and reasonable range, and mark the data that exceeds the normal and reasonable range as errors, including: numerical data exceeds the normal upper limit; code data exceeds the normal code range, etc. A3. For the pre-processed raw data, we need to collect complete data of a single ship during each voyage. Since the ship has records in different navigation service systems, a large amount of duplicate data will exist during the initial data collection process. These duplicate data show random fluctuations within a certain range at their respective nodes. This fluctuation is caused by differences in monitoring methods, monitoring time nodes, data storage accuracy, etc. of various navigation service systems. Each set of duplicate data only needs one set as the raw data for analysis. In order to retain the basic data characteristics of these data, the data is compressed through a clustering analysis algorithm.

Images