Marine equipment situation action associated knowledge graph construction method and system

Abstract

The invention relates to the technical field of data processing, in particular to a knowledge graph construction method and system associated with offshore equipment situation actions. Comprising the following steps: acquiring and processing offshore equipment action rules to obtain an action association relation matrix with a uniform format, and acquiring and processing offshore equipment situation elements to obtain situation element vectors; training the deep convolutional neural network model by using the action incidence relation matrix to obtain a network model structure and parameters; identifying action contents of the situation element vectors by utilizing a network model structure and parameters, and establishing a corresponding relationship between the situation element vectors and the action contents; utilizing a preset knowledge graph entity name to detect an inter-entity relationship in the corresponding relationship, and constructing a single entity model according to the situation element vector; and fusing the relationship between the entities and the single entity model to obtain a knowledge graph model. The problem of real-time construction of an accurate space-time dynamic mapping knowledge domain associated with situation actions in real-time situation data of offshore equipment is solved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method and system for constructing a knowledge graph related to maritime equipment situation and action. Background Art

[0002] In the prior art, the construction method of equipment knowledge graph mainly relies on the manual construction of expert knowledge. Specifically, expert knowledge is used to determine the entities, entity attributes and relationships between entities in the knowledge graph. This method has certain advantages in terms of accuracy, because experts can define various knowledge elements more accurately with their professional qualities and rich experience. However, its defects are also quite significant. On the one hand, expert knowledge itself may be redundant and conflicting, and the existing manual construction model is difficult to effectively identify and deal with these problems, which directly leads to a reduction in the effectiveness of knowledge graph construction. On the other hand, expert knowledge is often large in scale. In the process of manual construction, the huge workload makes it difficult to ensure that the knowledge graph construction can meet real-time requirements, especially when facing complex and changeable equipment application scenarios. The lack of real-time may delay key decisions and actions.

[0003] Advanced knowledge graph construction technology uses text data processing methods to automatically search and construct entities and relationships in the knowledge graph. This automated construction method is significantly more efficient than traditional manual methods, and can process large amounts of data and preliminarily construct a knowledge graph in a relatively short period of time. However, this technology is insufficient in terms of accuracy, and it is easy to have omissions in the process of entity and relationship construction, resulting in an incomplete knowledge graph. More importantly, this technology cannot organically integrate the real-time situation data of maritime equipment with expert knowledge rules, and it is difficult to achieve the construction of a knowledge graph that can be dynamically updated in time and space. For the data processing of maritime equipment command and control systems with extremely high real-time requirements, it is impossible to provide accurate, timely and effective knowledge graph support, which makes it difficult to meet the complex needs of maritime equipment operation and command and dispatch in actual applications, limiting the full utilization of equipment effectiveness and the efficient implementation of combat command. Summary of the invention

[0004] In order to solve the problem of being able to construct an accurate spatiotemporal dynamic knowledge graph of situation-action association in real-time in the real-time situation data of maritime equipment, the present invention provides a method and system for constructing a knowledge graph of situation-action association of maritime equipment.

[0005] In the first aspect, the present invention provides a method for constructing a knowledge graph associated with maritime equipment situation and action, which adopts the following technical solution: A method for constructing a knowledge graph of maritime equipment situation and action association, comprising: Acquire the maritime equipment action rules and process them to obtain an action correlation matrix in a unified format, acquire the maritime equipment situation elements and process them to obtain a situation element vector, wherein the situation element vector has the same attribute field distribution as the action correlation matrix; Construct a deep convolutional neural network model, and use the action association matrix to train the deep convolutional neural network model to obtain the network model structure and parameters; Using the network model structure and parameters to identify the action content of the situation element vector, and establish the corresponding relationship between the situation element vector and the action content; Detect the relationship between entities in the corresponding relationship using the preset knowledge graph entity name, where the relationship between entities includes the entity name and the action content vector, and construct a single entity model based on the situation element vector; The knowledge graph model is obtained by fusing the relationship between entities with the single entity model.

[0006] Furthermore, the processing to obtain a unified format action association matrix includes: Obtain the situational elements and action content of the maritime equipment action rules, and extract indivisible attributes; Constructing an action rule attribute set including multiple types, and numbering the attribute values ​​of the action rules in the action rule attribute set; The action rules in the action rule attribute set are supplemented with the full set of attributes of situation elements and action contents; the attribute values ​​of the action rules without description are set to zero, and the attribute values ​​of the remaining action rules are discretized; The completed action rules are queried and compared by the attributes of situation elements, and an action correlation matrix in a unified format is generated.

[0007] Furthermore, the action rule attribute set includes relative time series information of situation elements, target numbers of various types, quantity of targets of various types, relative spatial positions of targets with different numbers, energy endurance of targets with different numbers, attack power of targets with different numbers, target health of targets with different numbers, enemy and friendly situations of targets with different numbers, and action content of targets with different numbers.

[0008] Furthermore, the attribute values ​​of the action rules in the action rule attribute set are numbered, including selecting the attribute value with the smallest relative distance between the type target and the enemy target of the action rule as the smallest number in the type, and numbering them upward in sequence according to the increasing relative distance. If the distances between the corresponding type targets and the enemy targets are the same, they are randomly numbered.

[0009] Furthermore, the completed action rules are queried and compared based on the attributes of the situation elements, including comparing the attribute values ​​of the action content if the attribute values ​​of the situation elements are equal, retaining one of the rules if the attribute values ​​of the action content are equal, and determining them as conflicting rules if they are not equal, and deleting the rules.

[0010] Furthermore, the network model structure and parameters obtained by training the deep convolutional neural network model using the action association matrix include: Use the deep convolutional neural network model to perform forward reasoning on each row in the action association matrix, perform difference calculation and reverse reasoning on the attribute values ​​of the action content of each row in the action association matrix, and update the deep convolutional neural network parameters; The deep convolutional neural network model is trained for multiple rounds to obtain the network model structure and parameters.

[0011] Furthermore, the step of obtaining the situation element vector by processing the situation element of the marine equipment includes: Obtain the spatiotemporal situational elements of maritime equipment situational elements and extract indivisible attributes; Constructing a spatiotemporal situation element attribute set including multiple types, and numbering the attribute values ​​of the spatiotemporal situation elements in the spatiotemporal element attribute set; The attributes of all the spatiotemporal situation elements in the spatiotemporal element attribute set are completed, the attribute values ​​of the undescribed spatiotemporal situation elements are set to zero, and the attribute values ​​of the remaining spatiotemporal situations are discretized to generate a situation element vector, which has the same attribute field distribution as the action association relationship matrix.

[0012] Furthermore, the attribute set of spatiotemporal situation elements includes relative time series information of situation elements, numbers of various types of targets, numbers of various types of targets, relative spatial positions of targets with different numbers, energy endurance of targets with different numbers, attack power of targets with different numbers, target health values ​​of targets with different numbers, and enemy and friendly situations of targets with different numbers.

[0013] Furthermore, the attribute values ​​of the space-time situation elements in the space-time element attribute set are numbered, including selecting the attribute value with the smallest relative distance between the type target of the space-time situation element and the enemy target as the smallest number in the type, and numbering them upward in sequence according to the increasing relative distance. If the distances between the corresponding type targets and the enemy targets are the same, they are randomly numbered.

[0014] Furthermore, constructing a single entity model based on the situation element vector includes detecting entity information and other attribute information of the marine equipment in the situation element vector, constructing the entity information as a knowledge graph entity name, and constructing other attribute information as attribute information of the entity.

[0015] Furthermore, the knowledge graph model is obtained by integrating the relationship between entities and the single entity model, including: Obtain the relative time series information and type target number of the maritime equipment situation elements; According to the relative time series information and type target number, the relationship between entities is integrated with the single entity model to build a knowledge graph model in which multiple entities are interconnected; Intelligently update the knowledge graph model based on the relative time series information.

[0016] Second, a knowledge graph construction system for maritime equipment situation and action association includes: Data acquisition module, used to obtain maritime equipment action rules and maritime equipment situation elements; The relationship analysis module is connected with the data acquisition module and is used to process the action rules of the marine equipment and obtain the action correlation relationship matrix in a unified format; The spatiotemporal situation factor analysis module is connected to the data acquisition module and is used to process the situation factors of the marine equipment to obtain the situation factor vector; The relationship training module is connected to the relationship parsing module and is used to construct a deep convolutional neural network model. The action association relationship matrix is ​​used to train the deep convolutional neural network model to obtain the network model structure and parameters. The relationship preservation module is connected with the relationship training module and the spatiotemporal situation element analysis module, and is used to preserve the network model structure and parameters, identify the action content of the situation element vector using the network model structure and parameters, and establish the corresponding relationship between the situation element vector and the action content; A relationship output module, connected to the relationship storage module, is used to store the corresponding relationship between the situation element vector and the action content; A knowledge graph entity relationship parsing module, connected to the relationship output module, is used to detect the relationship between entities in the corresponding relationship using the preset knowledge graph entity name, wherein the relationship between entities includes the entity name and the action content vector; The knowledge graph preliminary construction module is connected to the spatiotemporal situation factor analysis module to construct a single entity model based on the situation factor vector; The knowledge graph intelligent update output module is connected to the knowledge graph preliminary construction module and the knowledge graph entity relationship parsing module at the same time, and is used to fuse the relationship between entities with the single entity model to obtain the knowledge graph model.

[0017] In a third aspect, the present invention provides a computer-readable storage medium storing a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device, and the method for constructing a knowledge graph associated with the situation and actions of maritime equipment is described.

[0018] In a fourth aspect, the present invention provides a terminal device comprising a processor and a computer-readable storage medium, the processor being used to implement various instructions; the computer-readable storage medium being used to store a plurality of instructions, wherein the instructions are suitable for being loaded by the processor and executing the method for constructing a knowledge graph associated with the situation and actions of maritime equipment.

[0019] In summary, the present invention has the following beneficial technical effects: 1. The present invention proposes a method and system for constructing a knowledge graph associated with maritime equipment situation and action, which can train the maritime equipment expert knowledge rules based on artificial intelligence algorithms to obtain a multi-layer neural network model that can intelligently reason about the dynamic relationship between entities. At the same time, the entity resolution algorithm is used to obtain entities, and the entities and the dynamic relationship between entities are combined to obtain a knowledge graph that is dynamically updated in time and space. Compared with the existing manual editing method of knowledge graphs, the efficiency is greatly improved. Compared with the method of intelligently extracting knowledge graphs from texts, the knowledge graph generated by the construction method has a standardized unified structure and the accuracy is greatly improved.

[0020] 2. The action correlation matrix generation method and the space-time situation element analysis method adopted in the present invention provide a unified description of the maritime equipment situation information affected by complex factors, and are conducive to the reasoning of artificial intelligence algorithms. It not only solves the problem of inconsistent structure of space-time situation elements and inability to effectively standardize processing, but also discretizes the data, thereby improving the accuracy of subsequent artificial intelligence algorithm training.

[0021] 3. The present invention constructs a single entity model based on the situation factor vector, constructs the entity information and other attribute information therein as the knowledge graph entity name and attribute information respectively, and obtains the knowledge graph model by detecting the relationship between entities in the corresponding relationship and integrating the relationship between entities and the single entity model. This process completely covers the construction process from data to entity and then to relationship, so that the constructed knowledge graph model comprehensively and accurately reflects the various situations of maritime equipment and the relationship between them, and can provide rich and accurate knowledge support for the command and control system of maritime equipment.

[0022] 4. Based on the relative time series information of the occurrence of maritime equipment situation elements, the knowledge graph model can be updated intelligently. The knowledge graph can adapt to the changing situation of maritime equipment in real time, always accurately reflect the current situation, greatly improve the scientificity and timeliness of decision-making, and help improve the overall operating efficiency and combat effectiveness of maritime equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flow chart of the method of embodiment 1 of the present invention; Figure 2 It is a schematic diagram of the system structure of Example 2 of the present invention. DETAILED DESCRIPTION

[0024] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0025] Example 1 Reference Figure 1 , a method for constructing a knowledge graph associated with maritime equipment situation and action in this embodiment includes: S1. Obtaining the maritime equipment action rules and processing them to obtain an action association matrix in a unified format, obtaining the maritime equipment situation elements and processing them to obtain a situation element vector, wherein the situation element vector has the same attribute field distribution as the action association matrix; This step achieves the initial integration and normalization of the original data, so that subsequent analysis, training, knowledge graph construction and other operations can be carried out based on a unified and orderly data structure, effectively avoiding processing difficulties and result deviations caused by inconsistent data formats or messy information.

[0026] The action association matrix in a unified format obtained by the processing includes: A1. Obtain the situation elements and action content of the maritime equipment action rules, and extract indivisible attributes; A2. Construct an action rule attribute set including multiple types, and number the attribute values ​​of the action rules in the action rule attribute set, including selecting the attribute value with the smallest relative distance between the target of the type of action rule and the enemy target as the smallest number in the type, and numbering them upward in sequence according to the increasing relative distance. If the distances between the target of the corresponding type and the enemy target are the same, they are randomly numbered.

[0027] The action rule attribute set includes relative time series information of situation elements, numbers of various types of targets, quantity of various types of targets, relative spatial positions of targets with different numbers, energy endurance of targets with different numbers, attack power of targets with different numbers, target health of targets with different numbers, enemy and friendly situations of targets with different numbers, and action contents of targets with different numbers.

[0028] A3 completes the attributes of the entire set of situation elements and action contents for the action rules in the action rule attribute set; the attribute values ​​of the action rules that are not described are set to zero, and the attribute values ​​of the remaining action rules are discretized; This ensures that each action rule has complete attribute information at the data level, facilitating unified analysis and comparison. Discretization helps convert continuous data into a discrete form that is more suitable for computer processing and model analysis, further optimizing data availability.

[0029] A4. Perform situation element attribute query and comparison on the completed action rules, and generate an action association matrix in a unified format, including comparing the attribute values ​​of the action content if the attribute values ​​of the situation elements are equal, retaining one of the rules if the attribute values ​​of the action content are equal, and deleting the rule if they are not equal. Finally, a unified action association matrix is ​​generated to ensure the consistency and accuracy of the data in the matrix, so that it can accurately reflect the internal relationship between the action rules of maritime equipment.

[0030] The step of obtaining the situation element vector of the maritime equipment by processing the situation element includes: B1. Obtain the spatiotemporal situational elements of the maritime equipment situational elements and extract indivisible attributes; B2. Construct a set of spatiotemporal situational element attributes including multiple types, and number the attribute values ​​of the spatiotemporal situation elements in the set of spatiotemporal element attributes, including selecting the attribute value with the smallest relative distance between the type target of the spatiotemporal situation element and the enemy target as the smallest number in the type, and numbering them upward in sequence according to the increasing relative distance. If the distances between the corresponding type targets and the enemy targets are the same, they are randomly numbered.

[0031] The attribute set of the spatiotemporal situational elements includes the relative time series information of the occurrence of the situational elements, the numbers of various types of targets, the number of various types of targets, the relative spatial positions of targets with different numbers, the energy endurance of targets with different numbers, the attack power of targets with different numbers, the target health values ​​of targets with different numbers, and the enemy and friendly situations of targets with different numbers.

[0032] B3. Complete the attributes of all the spatiotemporal situation elements in the spatiotemporal element attribute set, set the attribute values ​​of the undescribed spatiotemporal situation elements to zero, discretize the attribute values ​​of the remaining spatiotemporal situations, and generate a situation element vector. The situation element vector has the same attribute field distribution as the action association relationship matrix.

[0033] S2. Build a deep convolutional neural network model, and use the action association matrix to train the deep convolutional neural network model to obtain the network model structure and parameters, including: S21. Perform forward reasoning on each row in the action association matrix using a deep convolutional neural network model, perform difference calculation and reverse reasoning on the attribute values ​​of the action content of each row in the action association matrix, and update the deep convolutional neural network parameters; The deep convolutional neural network model performs forward reasoning on each row in the action association matrix, with the aim of allowing the model to extract key features and potential pattern information from the input situational element data based on the current network structure and parameter settings. Through the preliminary feature extraction and dimensionality reduction operation of the convolutional pooling layer of the convolutional neural network, the model can quickly capture the local features in the data. After forward reasoning, the difference calculation and reverse reasoning are performed on the attribute values ​​of the action content of each row in the action association matrix. The difference calculation can quantify the difference between the attribute values ​​of the action content predicted by the model and the actual action content, thereby providing key error information for reverse reasoning. Based on this error information, reverse reasoning calculates the gradient value of each neuron along the back propagation path of the network to determine the contribution of the network parameters to the error, thereby providing accurate adjustment direction and amplitude basis for updating the parameters of the deep convolutional neural network. Through this process, the model can continuously self-correct and optimize, gradually reduce the prediction error, and improve the ability to understand and grasp the action association relationship.

[0034] S22. Perform multiple rounds of training on the deep convolutional neural network model to obtain the network model structure and parameters. In this embodiment, 100 trainings are performed, and the specific design of the deep convolutional neural network model from front to back is: a vector composed of each row of situation factor data of the maritime equipment situation factor action association relationship matrix as input, a convolutional neural network convolution pooling layer, a residual network layer, a convolution pooling layer, a residual network layer, a convolution pooling layer, a convolutional neural network hidden layer, and a convolutional neural network output layer, wherein the residual network layer can effectively avoid the influence of unimportant attribute values ​​on the network model parameter training, and avoid the overfitting phenomenon of network model training.

[0035] The purpose of multiple rounds of training for the deep convolutional neural network model is to ensure that the model can fully learn the rich information contained in the action association matrix, so that it can achieve a stable and optimal performance state. As the number of training rounds increases, the model continuously adjusts parameters such as weights and biases in the network structure, and gradually converges to a state that can accurately describe the relationship between the action rules of maritime equipment and situation elements. During the 100 training processes set in this embodiment, the model gradually improves its prediction accuracy of action decisions under various maritime equipment situations, and finally obtains a network model structure and parameters with strong generalization ability, so that it can not only accurately fit the training data, but also make reasonable action content predictions for new and unseen maritime equipment situation element data.

[0036] S3. Using the network model structure and parameters to identify the action content of the situation element vector, and establish the corresponding relationship between the situation element vector and the action content; Based on the trained network model structure and parameters, the situation element vector is deeply analyzed and understood, so as to accurately identify the action content contained therein. In addition, a close correspondence between the situation element vector and the action content is further established, so that the situation information of the maritime equipment can be directly mapped to the corresponding action strategy or operation instruction. The establishment of this correspondence provides a key information bridge for the subsequent construction of the knowledge graph and the realization of intelligent decision-making of maritime equipment.

[0037] S4. Using the preset knowledge graph entity name to detect the entity relationship in the corresponding relationship, the entity relationship includes the entity name and the action content vector, and constructing a single entity model according to the situation element vector; The method of constructing a single entity model based on the situation element vector includes detecting entity information and other attribute information of the marine equipment in the situation element vector, constructing the entity information as a knowledge graph entity name, and constructing other attribute information as attribute information of the entity.

[0038] With the help of the preset knowledge graph entity names, we deeply analyze the previously established correspondence and dig out the relationship between entities, including clear entity names and the action content vectors associated with them. On this basis, we further build a single entity model based on the situation element vector to integrate and structure the information related to maritime equipment. This helps to clearly define the attributes and behaviors of each entity under the knowledge graph framework.

[0039] S5. The knowledge graph model is obtained by integrating the relationship between entities and the single entity model, including: S51. Obtain the relative time sequence information and type target number of the maritime equipment situation elements; S52. According to the occurrence relative time series information and the type target number, the relationship between entities is integrated with the single entity model to construct a knowledge graph model in which multiple entities are interconnected; By classifying different single entity models according to their corresponding type target numbers and determining the states and relationships of each entity at different time points based on the relative time series information, a knowledge graph model of multiple entities connected to each other is constructed. These knowledge graph models can fully display the associations between maritime equipment, their respective state changes, and the logical relationships of actions.

[0040] S53. Intelligently update the knowledge graph model based on the relative time series information.

[0041] Based on the relative time series information of the occurrence of maritime equipment situation elements, the intelligent update function of the knowledge graph model is realized. As the situation of maritime equipment continues to change, new situation element information is generated. This function can capture these changes in a timely manner and adjust and update the constructed knowledge graph model according to the new relative time series information. In this way, the knowledge graph model can always reflect the latest situation of maritime equipment, provide real-time and accurate knowledge support for application scenarios such as maritime equipment command and control systems, and ensure the timeliness and reliability of decision-making basis.

[0042] This embodiment provides a method for intelligently constructing a knowledge graph of maritime equipment situation-action association. The method is used in view of the fact that the existing equipment knowledge graph construction technology cannot be applied to the background that the expert knowledge rules are redundant, conflicting, and huge in number, and the expert knowledge and spatiotemporal situation data cannot be effectively associated. In order to be able to construct an accurate spatiotemporal dynamic knowledge graph of situation-action association in real time in the real-time situation data of maritime equipment, the expert knowledge rules are processed for redundancy and conflict through the situation element action association relationship matrix generation method, and then a large amount of expert knowledge is trained with a deep convolutional neural network model to obtain a situation element action association relationship classification model, and then the real-time spatiotemporal situation data is connected to the situation element action association relationship classification model to quickly obtain the relationship between knowledge graph entities, and multiple knowledge graph single entity models are obtained through the analysis of real-time spatiotemporal situation elements, and multiple knowledge graph single entity models and entity relationships are integrated to generate a knowledge graph of maritime equipment situation element-action association that is updated in real time according to the spatiotemporal situation data, which greatly improves the real-time and accuracy of knowledge graph construction. In addition, the use of this device and method is also to accurately and real-time construct intelligent knowledge graphs for massive real-time situation data and a large number of expert knowledge rules of a large number of marine equipment targets such as marine ships, unmanned ships, submersibles, buoys, etc. in the actual use of marine equipment. This can provide accurate, real-time and effective knowledge graph support for application systems such as command and control systems based on knowledge graphs, thereby improving the effectiveness and real-time performance of the marine equipment knowledge graph application system.

[0043] Example 2 Reference Figure 2 This embodiment provides a knowledge graph construction system for maritime equipment situation action association, including: Data acquisition module, used to obtain maritime equipment action rules and maritime equipment situation elements; The relationship analysis module is connected with the data acquisition module and is used to process the action rules of the marine equipment and obtain the action correlation relationship matrix in a unified format; The spatiotemporal situation factor analysis module is connected to the data acquisition module and is used to process the situation factors of the marine equipment to obtain the situation factor vector; The relationship training module is connected to the relationship parsing module and is used to construct a deep convolutional neural network model. The action association relationship matrix is ​​used to train the deep convolutional neural network model to obtain the network model structure and parameters. The relationship preservation module is connected with the relationship training module and the spatiotemporal situation element analysis module, and is used to preserve the network model structure and parameters, identify the action content of the situation element vector using the network model structure and parameters, and establish the corresponding relationship between the situation element vector and the action content; A relationship output module, connected to the relationship storage module, is used to store the corresponding relationship between the situation element vector and the action content; A knowledge graph entity relationship parsing module, connected to the relationship output module, is used to detect the relationship between entities in the corresponding relationship using the preset knowledge graph entity name, wherein the relationship between entities includes the entity name and the action content vector; The knowledge graph preliminary construction module is connected to the spatiotemporal situation factor analysis module to construct a single entity model based on the situation factor vector; The knowledge graph intelligent update output module is connected to the knowledge graph preliminary construction module and the knowledge graph entity relationship parsing module at the same time, and is used to fuse the relationship between entities with the single entity model to obtain the knowledge graph model.

[0044] A computer-readable storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device, and the method for constructing a knowledge graph associated with maritime equipment situation and action is disclosed.

[0045] A terminal device includes a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, wherein the instructions are suitable for being loaded by the processor and executing the method for constructing a knowledge graph associated with maritime equipment situation and action.

[0046] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for constructing a knowledge graph of maritime equipment situation and action association, characterized in that: include: Acquire the maritime equipment action rules and process them to obtain an action correlation matrix in a unified format, acquire the maritime equipment situation elements and process them to obtain a situation element vector, wherein the situation element vector has the same attribute field distribution as the action correlation matrix; Construct a deep convolutional neural network model, and use the action association matrix to train the deep convolutional neural network model to obtain the network model structure and parameters; Using the network model structure and parameters to identify the action content of the situation element vector, and establish the corresponding relationship between the situation element vector and the action content; Detecting the relationship between entities in the corresponding relationship using the preset knowledge graph entity name, wherein the relationship between entities includes the entity name and the action content vector, and constructing a single entity model according to the situation element vector; The knowledge graph model is obtained by fusing the relationship between entities with the single entity model.

2. The method for constructing a knowledge graph of maritime equipment situation and action association according to claim 1, characterized in that: The action association matrix in a unified format obtained by the processing includes: Obtain the situational elements and action content of the maritime equipment action rules, and extract indivisible attributes; Constructing an action rule attribute set including multiple types, and numbering the attribute values ​​of the action rules in the action rule attribute set; The action rules in the action rule attribute set are supplemented with the full set of attributes of situation elements and action contents; the attribute values ​​of the action rules without description are set to zero, and the attribute values ​​of the remaining action rules are discretized; The completed action rules are queried and compared by the attributes of situation elements, and an action correlation matrix in a unified format is generated.

3. The method for constructing a knowledge graph of maritime equipment situation and action association according to claim 2, characterized in that: The attribute values ​​of the action rules in the action rule attribute set are numbered, including selecting the attribute value with the smallest relative distance between the type target and the enemy target of the action rule as the smallest number in the type, and numbering them upward in sequence according to the increasing relative distance. If the distances between the corresponding type target and the enemy target are the same, they are randomly numbered.

4. The method for constructing a knowledge graph of maritime equipment situation and action association according to claim 2, characterized in that: The completed action rules are queried and compared with the attributes of the situation elements, including comparing the attribute values ​​of the action content if the attribute values ​​of the situation elements are equal, retaining one of the rules if the attribute values ​​of the action content are equal, and determining them as conflicting rules if they are not equal, and deleting the rules.

5. The method for constructing a knowledge graph of maritime equipment situation and action association according to claim 1, characterized in that: The network model structure and parameters obtained by training the deep convolutional neural network model using the action association matrix include: Use the deep convolutional neural network model to perform forward reasoning on each row in the action association matrix, perform difference calculation and reverse reasoning on the attribute values ​​of the action content of each row in the action association matrix, and update the deep convolutional neural network parameters; The deep convolutional neural network model is trained for multiple rounds to obtain the network model structure and parameters.

6. The method for constructing a knowledge graph of maritime equipment situation and action association according to claim 1, characterized in that: The step of obtaining the situation element vector of the marine equipment situation element comprises: Obtain the spatiotemporal situational elements of maritime equipment situational elements and extract indivisible attributes; Constructing a spatiotemporal situation element attribute set including multiple types, and numbering the attribute values ​​of the spatiotemporal situation elements in the spatiotemporal element attribute set; The attributes of all the spatiotemporal situation elements in the spatiotemporal element attribute set are completed, the attribute values ​​of the undescribed spatiotemporal situation elements are set to zero, and the attribute values ​​of the remaining spatiotemporal situations are discretized to generate a situation element vector, which has the same attribute field distribution as the action association relationship matrix.

7. The method for constructing a knowledge graph of maritime equipment situation and action association according to claim 6, characterized in that: The attribute values ​​of the space-time situation elements in the space-time element attribute set are numbered, including selecting the attribute value with the smallest relative distance between the type target of the space-time situation element and the enemy target as the smallest number in the type, and numbering them upward in sequence according to the increasing relative distance. If the distances between the corresponding type targets and the enemy targets are the same, they are randomly numbered.

8. The method for constructing a knowledge graph of maritime equipment situation and action association according to claim 1, characterized in that: The method of constructing a single entity model based on the situation element vector includes detecting entity information and other attribute information of the marine equipment in the situation element vector, constructing the entity information as a knowledge graph entity name, and constructing other attribute information as attribute information of the entity.

9. The method for constructing a knowledge graph of maritime equipment situation and action association according to claim 1, characterized in that: The knowledge graph model is obtained by integrating the relationship between entities and the single entity model, including: Obtain the relative time series information and type target number of the maritime equipment situation elements; According to the relative time series information and type target number, the relationship between entities is integrated with the single entity model to build a knowledge graph model in which multiple entities are interconnected; Intelligently update the knowledge graph model based on the relative time series information.

10. A knowledge graph construction system for maritime equipment situation and action association, comprising: Data acquisition module, used to obtain maritime equipment action rules and maritime equipment situation elements; The relationship analysis module is connected with the data acquisition module and is used to process the action rules of the marine equipment and obtain the action correlation relationship matrix in a unified format; The spatiotemporal situation factor analysis module is connected to the data acquisition module and is used to process the situation factors of the marine equipment to obtain the situation factor vector; The relationship training module is connected to the relationship parsing module and is used to construct a deep convolutional neural network model. The action association relationship matrix is ​​used to train the deep convolutional neural network model to obtain the network model structure and parameters. The relationship preservation module is connected with the relationship training module and the spatiotemporal situation element analysis module, and is used to preserve the network model structure and parameters, identify the action content of the situation element vector using the network model structure and parameters, and establish the corresponding relationship between the situation element vector and the action content; A relationship output module, connected to the relationship storage module, is used to store the corresponding relationship between the situation element vector and the action content; A knowledge graph entity relationship parsing module, connected to the relationship output module, is used to detect the relationship between entities in the corresponding relationship using the preset knowledge graph entity name, wherein the relationship between entities includes the entity name and the action content vector; The knowledge graph preliminary construction module is connected to the spatiotemporal situation factor analysis module to construct a single entity model based on the situation factor vector; The knowledge graph intelligent update output module is connected to the knowledge graph preliminary construction module and the knowledge graph entity relationship parsing module at the same time, and is used to fuse the relationship between entities with the single entity model to obtain the knowledge graph model.

Images