Combat simulation system and method based on knowledge graph and Bayesian network

Abstract

The simulation combat simulation system and method based on knowledge graph and Bayesian network, the data acquisition unit collects relevant data of simulation combat resources through multiple data collectors to obtain simulation combat resource data; the data processing unit performs data processing on the collected simulation combat resource data to obtain normalized data; the knowledge graph construction unit constructs the knowledge graph according to the normalized data to generate a structured graph of simulation combat resources; the Bayesian network construction unit obtains a Bayesian network model according to the nodes and node relationships of the structured graph of simulation combat resources; the result evaluation unit performs dynamic reasoning of simulation combat simulation according to the Bayesian network model, and generates a simulation report. The present invention comprehensively understands the detection capability, synergy efficiency and overall detection effect of each data collector under different conditions, performs simulation combat simulation and generates corresponding simulation reports, thereby providing a scientific basis for command decision-making.

Description

Technical Field

[0001] The present invention relates to the technical field of simulation combat resource capability assessment, and in particular to a simulation combat simulation system and method based on knowledge graph and Bayesian network. Background Art

[0002] With the complexity and high technology of modern warfare, higher requirements are placed on the management and evaluation of simulation combat resources. The current simulation combat resource evaluation methods mainly include expert evaluation method, mathematical analysis method and computer simulation analysis method. These methods have certain effects in the simulation combat resource evaluation of simple systems, but have limitations when facing complex systems. Traditional simulation combat resource evaluation methods are difficult to meet the needs of multi-dimensional and multi-granular simulation combat resource evaluation. Therefore, a new evaluation method is urgently needed to improve the accuracy and comprehensiveness of simulation combat resource evaluation.

[0003] In modern warfare, the management and evaluation of combat resources are important links in command decision-making. Traditional simulation combat resource evaluation methods rely on expert experience and mathematical models, which are incapable of coping with complex and dynamically changing combat environments. For example, the expert evaluation method is greatly affected by the subjective judgment of experts, and it is difficult to ensure the objectivity and consistency of the evaluation results; the mathematical analysis method requires a lot of prior knowledge and accurate data support, and the calculation complexity is high when dealing with complex systems; although the computer simulation analysis method can simulate complex environments, it is limited by the accuracy of the model and the completeness of the data, and the reliability and authenticity of the simulation combat resource evaluation results still need to be further verified.

[0004] Therefore, the problems existing in the prior art need to be further improved and developed. Summary of the invention

[0005] (I) Purpose of the invention: In order to solve the problems existing in the above-mentioned prior art, the purpose of the present invention is to provide a combat simulation system and method based on knowledge graph and Bayesian network, which realizes comprehensive and scientific simulation evaluation in complex combat environment by constructing a combat simulation knowledge graph and combining Bayesian network to perform effectiveness result evaluation reasoning.

[0006] (II) Technical solution: In order to solve the above technical problems, this technical solution provides a combat simulation system based on knowledge graph and Bayesian network, including a data acquisition unit, a data processing unit, a knowledge graph construction unit, a Bayesian network construction unit and a result evaluation unit; the data acquisition unit includes a plurality of data collectors, and the relevant data of the simulated combat resources are collected through the plurality of data collectors to obtain the simulated combat resource data;

[0007] The data processing unit processes the collected simulated combat resource data to obtain normalized data; the data processing includes data cleaning, format conversion and normalization processing;

[0008] The knowledge graph construction unit constructs a knowledge graph according to the normalized data to generate a structured graph of simulated combat resources; the knowledge graph construction unit includes a node determination module and a relationship determination module, the node determination module identifies and defines the data collector and its attributes, and the relationship determination module analyzes the synergy between the various data collectors to determine the relationship between the nodes;

[0009] The Bayesian network construction unit obtains a Bayesian network model according to the nodes and node relationships of the simulated combat resource structured graph; initializes the nodes of the Bayesian network model according to the simulated combat resource structured graph; then constructs a conditional probability table corresponding to each node in the Bayesian network model based on statistical analysis and actual data; optimizes the structure of the Bayesian network model, verifies the Bayesian network model, and evaluates the performance indicators of the Bayesian network model;

[0010] The result evaluation unit performs dynamic reasoning of the simulated combat simulation according to the Bayesian network model and generates a simulation report.

[0011] Preferably, the data cleaning refers to calculating the mean μ and standard deviation σ of the data points in the data set.

[0012] , ,

[0013] Where N is the total number of data points in the dataset, X i represents the i-th data point.

[0014] Then calculate the Z-Score value of each data point.

[0015] ,

[0016] Finally, an abnormal threshold is set, and data points whose absolute value of Z-Score is greater than the abnormal threshold are regarded as abnormal data and removed.

[0017] Preferably, the normalization process adopts Min-Max normalization or Z-Score standardization technology. When the type of the simulated combat resource data is in JSON format, Min-Max normalization is selected; when the type of the simulated combat resource data is in CSV format, Z-Score standardization is selected to scale the simulated combat resource data to the same dimension and range. ,

[0018] ,

[0019] Where X represents the data point; X min Indicates the minimum value among the data points; X max Indicates the maximum value among the data points; X norm represents the normalized data point; μ represents the mean of the data set; σ represents the standard deviation of the data set.

[0020] Preferably, each node represents a data collector, and the attributes of each node are defined according to the data collector corresponding to the node.

[0021] Preferably, the relationship determination module determines the direction and weight of the edge by calculating the comprehensive weight W. The directed edge represents the direction of the relationship between nodes, and the weight represents the strength of the relationship, including the data transmission success rate S (%), the command response time T, the environmental impact coefficient E (ms) and the redundancy level coefficient R.

[0022] ,

[0023] Among them, α, β, γ, and δ represent the weight coefficients of each factor, respectively, satisfying α+β+γ+δ=1.

[0024] Preferably, each node has a conditional probability table, which represents the probability distribution of the node under different conditions; for nodes without parent nodes, the prior probability is directly calculated.

[0025] ,

[0026] Among them, N(X i =x) represents node X i The number of nodes in state x; N represents the number of nodes X i Total number of

[0027] For nodes with parent nodes, the conditional probability is calculated based on the state of its parent node and the actual data.

[0028] ,

[0029] Among them, X i Represents the current node; x represents a state of the current node; p represents the state combination of the parent node.

[0030] Preferably, cross-validation is used to verify the Bayesian network model, the data set is divided into a training set and a test set, the Bayesian network model is trained using the training set, and the performance indicators of the Bayesian network model on different data sets are evaluated on the test set, the structure and conditional probability table of the Bayesian network are adjusted, and the Bayesian network model is continuously optimized.

[0031] Preferably, the performance indicator evaluation includes accuracy, recall and F1 score.

[0032] Evaluate the accuracy of the Bayesian network model:

[0033] ,

[0034] Among them, TP represents the number of positive classes that are actually predicted to be positive; TN represents the number of negative classes that are actually predicted to be negative; FP represents the number of negative classes that are actually predicted to be positive; FN represents the number of positive classes that are actually predicted to be negative.

[0035] Evaluate the recall of the Bayesian network model, that is, the proportion of correct predictions in actual positive examples:

[0036] ,

[0037] Among them, TP represents the number of positive classes that are actually predicted to be positive; FN represents the number of negative classes that are actually predicted to be positive;

[0038] Evaluate the F1 score of the Bayesian network model, which is the harmonic mean of precision and recall:

[0039] .

[0040] Preferably, analyze the impact of each node in the Bayesian network on the overall performance, identify key nodes and weak links, and evaluate the marginal impact of each node on the target node:

[0041] ,

[0042] Among them, Impact(X i ) represents node X i The degree of influence on the target node; States(X i ) represents node X i All possible states of P(Target|X i =x) means the target node is at X i The probability of being in state x;

[0043] By calculating the degree of influence of the conditional probability changes between nodes on the probability distribution of the target node, the structure and conditional probability table of the Bayesian network model are adjusted to continuously optimize the performance of the Bayesian network model.

[0044] The simulation combat simulation method based on knowledge graph and Bayesian network is applicable to the simulation combat simulation system based on knowledge graph and Bayesian network, and specifically includes the following steps:

[0045] Step 1: The data acquisition unit collects relevant data of the simulated combat resources through a variety of data collectors to obtain simulated combat resource data;

[0046] Step 2: The data processing unit processes the collected simulation combat resource data to obtain normalized data; the data processing includes data cleaning, format conversion and normalization processing;

[0047] Step 3: The knowledge graph construction unit constructs a knowledge graph according to the normalized data to generate a structured graph of simulated combat resources; the knowledge graph construction unit includes a node determination module and a relationship determination module, the node determination module identifies and defines the data collector and its attributes, and the relationship determination module analyzes the synergy between the data collectors and determines the relationship between the nodes;

[0048] Step 4, the Bayesian network construction unit obtains a Bayesian network model according to the nodes and node relationships of the simulated combat resource structured graph; initializes the nodes of the Bayesian network model according to the simulated combat resource structured graph; then constructs a conditional probability table corresponding to each node in the Bayesian network model based on statistical analysis and actual data; optimizes the structure of the Bayesian network model, verifies the Bayesian network model, and evaluates the performance indicators of the Bayesian network model;

[0049] Step 5: The result evaluation unit performs dynamic reasoning of the simulated combat simulation based on the Bayesian network model and generates a simulation report.

[0050] (III) Beneficial effects: The combat simulation system and method based on knowledge graph and Bayesian network of the present invention can comprehensively understand the detection capability, synergy efficiency and overall detection effect of each data collector under different conditions, conduct combat simulation and generate corresponding simulation reports, thereby providing a scientific basis for command decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a structural diagram of a combat simulation system based on knowledge graph and Bayesian network of the present invention;

[0052] Figure 2 It is a step diagram of the combat simulation method based on knowledge graph and Bayesian network of the present invention;

[0053] Figure 3 is a schematic diagram of a process of using a radar sensor as a data collector in the first embodiment of the present invention;

[0054] Figure 4 It is a schematic diagram of the structure of the simulated combat resource structured map with the radar sensor as a data collector;

[0055] Figure 5It is a schematic diagram of the structure of the Bayesian network model of the radar sensor as a data collector;

[0056] Figure 6 It is a flow chart of the second embodiment of the present invention in which the UAV is used as a data collector. DETAILED DESCRIPTION

[0057] The present invention is further described in detail below in conjunction with preferred embodiments. More details are elaborated in the following description to facilitate a full understanding of the present invention. However, the present invention can obviously be implemented in a variety of other ways different from the description. Those skilled in the art can make similar generalizations and deductions based on actual application situations without violating the connotation of the present invention. Therefore, the protection scope of the present invention should not be limited by the content of this specific embodiment.

[0058] The accompanying drawings are schematic diagrams of embodiments of the present invention. It should be noted that the drawings are only examples and are not drawn to scale, and should not be used to limit the actual protection scope of the present invention.

[0059] A simulation combat system based on knowledge graph and Bayesian network, such as Figure 1 As shown, it includes a data acquisition unit, a data processing unit, a knowledge graph construction unit, a Bayesian network construction unit and a result evaluation unit. The data acquisition unit includes a plurality of data collectors, through which the relevant data of the simulated combat resources are collected to obtain the simulated combat resource data. The data processing unit performs data processing on the collected simulated combat resource data to obtain normalized data. The knowledge graph construction unit constructs a knowledge graph according to the normalized data and generates a structured graph of simulated combat resources. The Bayesian network construction unit obtains a Bayesian network model according to the nodes and node relationships of the structured graph of simulated combat resources. The result evaluation unit performs dynamic reasoning of simulated combat simulation according to the Bayesian network model and generates a simulation report.

[0060] The simulation combat simulation method based on knowledge graph and Bayesian network is applicable to the simulation combat simulation system based on knowledge graph and Bayesian network, such as Figure 2 As shown, the specific steps include:

[0061] Step 1: The data acquisition unit collects relevant data of the simulated combat resources through a variety of data collectors to obtain simulated combat resource data.

[0062] Step 2: The data processing unit processes the collected simulation combat resource data to obtain normalized data.

[0063] Step 3: The knowledge graph construction unit constructs a knowledge graph based on the normalized data to generate a structured graph of simulated combat resources.

[0064] Step 4: The Bayesian network construction unit obtains a Bayesian network model according to the nodes and node relationships of the structured graph of simulated combat resources.

[0065] Step 5: The result evaluation unit performs dynamic reasoning of the simulated combat simulation based on the Bayesian network model and generates a simulation report.

[0066] More specifically, the specific operation of the data acquisition unit in step 1 is:

[0067] The main task of the data acquisition unit is to collect simulated combat resource data through a variety of data collectors, including actual combat drill data, simulated combat simulation data and historical record data. The simulated combat simulation data refers to the actual situation of the data collector in a multi-target environment simulated by high-fidelity simulation software. The historical record data includes data from previous combat drills and / or actual combat. These simulated combat resource data are rich in sources, and the comprehensiveness and representativeness of the simulated combat resource data are ensured by combining actual combat drill data with simulated combat simulation data and historical record data.

[0068] Step 101: Initialize the configuration of the data collector.

[0069] The data collector refers to an IoT data collection device, including reconnaissance and detection equipment, monitoring equipment and data recording equipment. The data collector can acquire and store dynamic information of simulated combat resources in real time, providing a basis for subsequent processing and analysis. The reconnaissance and detection equipment includes but is not limited to satellite sensors, ground radars, etc.; the monitoring equipment includes command and control center monitoring equipment, communication link monitoring equipment, environmental monitoring equipment and maintenance station monitoring equipment, etc., to monitor the status of simulated combat resource data in real time; the data recording equipment is used to safely store and efficiently manage the large amount of collected simulated combat resource data.

[0070] Step 102: Use the data collector to collect relevant data of simulated combat resources in real time.

[0071] Collect detection data, such as target location, detection range, etc. from reconnaissance detection equipment. Collect control instructions and feedback data from command and control center monitoring equipment. Collect data transmission status from communication link monitoring equipment. Collect weather and terrain data from environmental monitoring equipment. Collect equipment maintenance records and status data from maintenance station monitoring equipment.

[0072] Step 103: Transmission and storage of simulation combat resource data to a data set.

[0073] The collected simulated combat resource data is transmitted to the command and control center in real time through a high-frequency data transmission link. A data recording device is used to store all collected simulated combat resource data in a data set to ensure the timeliness and integrity of the simulated combat resource data.

[0074] More specifically, the specific operations of the data processing unit in step 2 are:

[0075] The data processing unit processes the simulated combat resource data collected by the data collector and outputs normalized data in a unified format to ensure the accuracy, consistency and comparability of the normalized data.

[0076] The data processing includes data cleaning, format conversion and normalization. The data cleaning refers to removing noise and abnormal data. The present invention uses the Z-Score method to detect abnormal data, as shown in formula (1), to automatically identify and correct abnormal data, and ensure the accuracy and reliability of the simulated combat resource data.

[0077] (1),

[0078] Among them, Z represents the Z-Score value; X represents the data point; μ represents the mean of the data set; σ represents the standard deviation of the data set.

[0079] The specific steps of data cleaning are:

[0080] Step a1: Calculate the mean μ and standard deviation σ of the data points in the data set.

[0081] , ,

[0082] Where N is the total number of data points in the dataset, X i represents the i-th data point.

[0083] Step a2: Calculate the Z-Score value of each data point.

[0084] .

[0085] Step a3: set an abnormal threshold, and regard data points whose absolute value of Z-Score is greater than the abnormal threshold as abnormal data and remove them.

[0086] The abnormal threshold can be selected according to the actual situation, and is usually set to 3. That is, in the present invention, the absolute value of the Z-Score | Z i The data points with values ​​greater than 3 are cleaned and removed.

[0087] The format conversion standardizes the simulated combat resource data from different sources into a unified format through a data parsing tool, which is convenient for subsequent analysis and processing. The format of the data after the format conversion includes but is not limited to JSON or CSV, etc., and the present invention includes both JSON and CSV formats.

[0088] The normalization process uses standardization or normalization to scale the simulated combat resource data to the same range, ensuring that the simulated combat resource data collected by different data collectors have consistent dimensions and ranges, thereby improving the consistency and comparability of the simulated combat resource data.

[0089] The specific steps of the normalization process are:

[0090] Step c1: Select a normalization method according to the type of simulation combat resource data.

[0091] The normalization process adopts Min-Max normalization or Z-Score standardization technology. When the type of the simulation combat resource data is in JSON format, Min-Max normalization is selected, as shown in formula (2); when the type of the simulation combat resource data is in CSV format, Z-Score standardization is selected to scale the simulation combat resource data to the same dimension and range, as shown in formula (3).

[0092] (2),

[0093] (3),

[0094] Where X represents the data point; X min Indicates the minimum value among the data points; X max represents the maximum value of the data point; μ represents the mean of the data set; σ represents the standard deviation of the data set.

[0095] Step c2: Apply the normalization method to normalize the simulated combat resource data and scale them to the same dimension and range.

[0096] More specifically, the specific operations of the knowledge graph construction unit in step 3 are:

[0097] The knowledge graph construction unit includes a node determination module and a relationship determination module. The nodes and node relationships are determined according to the normalized data, so as to construct a knowledge graph reflecting the data collectors and their mutual relationships in the simulation combat simulation system, obtain a structured graph of simulation combat resources, and provide a basis for the construction of the Bayesian network model.

[0098] The knowledge graph is a technical problem that uses a graph model to describe knowledge and model the relationship between all things in the world. The knowledge graph consists of nodes and edges. Nodes can be entities or abstract concepts; edges can be relationships between entities.

[0099] Step 301: The node determination module identifies and defines a data collector and its attributes.

[0100] Each node represents a data collector. The attributes of the node are the information of the corresponding node, including the technical parameters, functional characteristics and operating status of the corresponding data collector. The attributes of each node are defined according to the technical parameters, functional characteristics, operating status redundancy level and environmental impact of the data collector corresponding to the node.

[0101] It should be noted that the determination of nodes requires comprehensive consideration of the types, functions and roles of data collectors. For example, satellite nodes can be further subdivided into different types of reconnaissance satellites, communication satellites, etc., and each type of satellite plays a different role in the mission. The determination of node attributes requires a combination of actual data and expert knowledge to ensure that all key performance indicators of the data collector are covered.

[0102] Step 302: The relationship determination module analyzes the synergy between the data collectors to determine the relationship between the nodes.

[0103] The synergy needs to consider the coordination relationship of data collectors in the task, use statistical methods to calculate the degree of mutual influence between data collectors, and determine the direction and weight of the edge. The directed edge represents the direction of the relationship between nodes, and the weight represents the strength of the relationship.

[0104] More specifically, through data analysis and expert rules, the specific relationship between entities is extracted from the data, and the relationship weight is calculated based on the multi-index comprehensive evaluation model to reflect the synergy and the degree of mutual influence. As shown in formula (3), the calculation of the comprehensive weight W takes into account multiple factors, including data transmission success rate S (%), command response time T, environmental impact coefficient E (ms) and redundancy level coefficient R.

[0105] (3),

[0106] Among them, α, β, γ, and δ represent the weight coefficients of each factor, respectively, satisfying α+β+γ+δ=1.

[0107] The statistical methods may include correlation analysis, regression analysis, etc.

[0108] Step 303: Obtain a structured graph of simulated combat resources based on the nodes and node relationships.

[0109] The nodes and node relationships are stored in a graph database, and the structure of the knowledge graph is analyzed through graph theory algorithms to identify key nodes and important relationships, eliminate redundant relationships, and optimize the structured graph of simulated combat resources to ensure the accuracy and efficiency of the structured graph of simulated combat resources. In addition, a graph neural network is introduced for intelligent structural optimization, automatic identification and adjustment of key nodes and relationships, and enhancement of the overall effectiveness of the structured graph of simulated combat resources.

[0110] More specifically, the specific operation of the Bayesian network construction unit in step 4 is:

[0111] The Bayesian network model is constructed by converting the nodes and node relationships in the structured graph of the simulated combat resources into nodes in the Bayesian network model and optimizing the Bayesian network model.

[0112] The Bayesian network model is a probabilistic graph model, which is constructed on the basis of the structured graph of simulated combat resources. It represents the influencing factors and their relationships in the simulated combat simulation system through nodes and directed edges. The nodes represent random variables, and the edges represent conditional probability relationships. The Bayesian network model evaluates the overall results of the simulated combat simulation system through the calculation of conditional probabilities, and can effectively deal with uncertainty problems.

[0113] Step 401: Initialize the nodes of the Bayesian network model according to the structured graph of simulated combat resources.

[0114] The entity nodes in the knowledge graph are mapped to random variable nodes in the Bayesian network, and the entity attributes are mapped to the states or attributes of the nodes. Each node represents a random variable, and the initial state of the node of the Bayesian network model is set according to the node attributes in the structured graph of the simulated combat resources.

[0115] Step 402: construct a conditional probability table corresponding to each node in the Bayesian network model based on statistical analysis and actual data.

[0116] Each node has a conditional probability table (CPT), which represents the probability distribution of the node under different conditions. The construction of the conditional probability table needs to be based on actual data and statistical methods, using historical data and real-time data, using maximum likelihood estimation (MLE) or Bayesian estimation methods, calculating the probability value under each condition, and adjusting it in combination with expert knowledge to ensure the accuracy and robustness of the probability distribution.

[0117] For nodes without parent nodes, the prior probability is calculated directly, as shown in formula (4).

[0118] (4),

[0119] Among them, N(Xi =x) represents node X i The number of nodes in state x; N represents the number of nodes X i The total quantity.

[0120] For nodes with parent nodes, the conditional probability is calculated based on the state of its parent node and the actual data, as shown in formula (5).

[0121] (5),

[0122] Among them, X i Represents the current node; x represents a state of the current node; p represents the state combination of the parent node.

[0123] When the data is insufficient or the data quality is not high, the conditional probability table can be adjusted and supplemented in combination with expert knowledge or experience.

[0124] Step 403: Optimize the structure of the Bayesian network model.

[0125] The optimization of the Bayesian network model structure can be to automatically learn the structure of the Bayesian network using a structural learning algorithm, and determine the connection relationship between nodes based on the structured map of simulated combat resources and actual data. It can also be through sensitivity analysis or information gain calculation to identify and delete redundant connections that have little effect on the reasoning results. Decompose the complex network into multiple sub-networks or hierarchical structures to improve the reasoning efficiency. Ensure that the connectivity and dependency of the network are reasonable, avoid information loss or error propagation caused by optimization, and ensure the rationality and efficiency of the Bayesian network model structure.

[0126] The structure learning algorithm includes but is not limited to greedy search, Bayesian scoring, etc. The specific operations are:

[0127] Greedy search is used to identify and optimize the connection relationship between nodes in the Bayesian network model, reflecting the dependence and influence between nodes. Starting from the initial structure, the greedy search gradually tries to add or delete edges in the Bayesian network model to maximize the Bayesian score. It is found that the connection between some reconnaissance equipment and communication links has little impact on the overall evaluation effect, so these redundant connections are removed to simplify the network structure.

[0128] The Bayesian score is a universal scoring function method that can be directly used to evaluate and compare different Bayesian network structures. It helps the structural learning algorithm select the optimal network structure by comprehensively considering the model's fit (likelihood) and complexity (prior probability).

[0129] The specific operations of decomposing a complex network into multiple sub-networks or hierarchical structures are: adopting functional grouping, grouping based on coupling degree, and grouping rules of hierarchical structure. The functional grouping refers to grouping nodes in the network according to function or role, including operation layer, control layer, support layer, etc. The coupling-based grouping refers to grouping highly coupled or highly correlated (similar attribute states) nodes into one group to reduce the dependency and interaction complexity between sub-networks. The coupling degree can be measured by indicators such as conditional dependency and statistical correlation. The hierarchical structure divides the network into different levels according to the direction of data flow or control flow. Usually, the bottom layer is responsible for data collection and processing, the middle layer is responsible for data transmission and coordination, and the upper layer is responsible for decision-making and command. The operation layer transmits data to the control layer through a communication link, and the control layer relies on the environment and maintenance information provided by the support layer.

[0130] Step 404: verify the Bayesian network model and evaluate the performance indicators of the Bayesian network model.

[0131] Cross-validation is used in the present invention. Specifically, the data set is divided into a training set (80%) and a test set (20%), the Bayesian network model is trained using the training set, and the performance indicators of the Bayesian network model on different data sets are evaluated on the test set, key nodes and weak links are identified, the marginal influence of each node on the target node is evaluated, the structure and conditional probability table of the Bayesian network are adjusted, and the Bayesian network model is continuously optimized.

[0132] The performance indicators of the verified Bayesian network model are evaluated, including accuracy, recall rate and F1 score; as shown in formula (6), the accuracy of the Bayesian network model is evaluated.

[0133] (6),

[0134] Among them, TP represents the number of positive classes that are actually predicted to be positive; TN represents the number of negative classes that are actually predicted to be negative; FP represents the number of negative classes that are actually predicted to be positive; and FN represents the number of positive classes that are actually predicted to be negative.

[0135] The recall rate of the Bayesian network model is evaluated as shown in formula (7), that is, the proportion of correct predictions in actual positive examples.

[0136] (7),

[0137] Among them, TP represents the number of true positives and predicted positives; FN represents the number of true positives and predicted negatives.

[0138] As shown in formula (8), the F1 score of the Bayesian network model is evaluated, which is the harmonic mean of precision and recall.

[0139] (8).

[0140] Analyze the impact of each node in the Bayesian network on the overall performance, identify key nodes and weak links, and evaluate the marginal influence of each node on the target node. As shown in formula (9), the influence of the conditional probability change between nodes on the probability distribution of the target node is calculated. According to the final verification results, adjust the structure and conditional probability table of the Bayesian network model to continuously optimize the performance of the Bayesian network model.

[0141] (9),

[0142] Among them, Impact(X i ) represents node X i The degree of influence on the target node; States(X i ) represents node X i All possible states of P(Target|X i =x) means the target node is at X i The probability of being in state x.

[0143] More specifically, the specific operations of the result evaluation unit in step 5 are:

[0144] The Bayesian network model is used to perform conditional probability calculations and generate a detailed simulation report. First, the simulated combat resource data is input into the Bayesian network model, including the operating status of the data collector and the mission environment. Then, based on the Bayesian theorem and the chain rule, the conditional probability of each node is calculated to ensure the accuracy and efficiency of the calculation results. Finally, based on the calculation results of the conditional probability, the simulation combat simulation system dynamically infers the results of the simulation combat simulation and generates a simulation report including key indicators such as detection success rate and detection coverage, providing a scientific basis for command decision-making.

[0145] Step 501: Input the simulated combat resource data into the Bayesian network model.

[0146] The simulation combat resource data includes sensor detection data, target tracking data, communication link status, environmental data and maintenance records, etc.

[0147] Step 502: The Bayesian network model performs conditional probability calculation based on the input simulation combat resource data.

[0148] Determine the target node, as shown in formula (10), and use the conditional probability table in the Bayesian network to calculate the posterior probability of the target node based on the simulated combat resource data. Then, use the chain rule to calculate the probability of each node layer by layer, as shown in formula (11), and finally obtain the comprehensive performance index of the target node.

[0149] (10),

[0150] Among them, P(X|Y) represents the probability of event X occurring under evidence Y; P(Y|X) represents the probability of evidence Y occurring under the condition that event X occurs; P(X) represents the prior probability of event X; and P(Y) represents the total probability of evidence Y.

[0151] (11),

[0152] Among them, P(A,B∣C) is the probability of A+B occurring under the premise that C occurs; P(A∣B,C) is the probability of A occurring under the premise that B+C occurs; P(B∣C) is the probability of B occurring under the premise that C occurs.

[0153] Step 503: Generate a simulation report including key indicators such as detection success rate, information transmission success rate, system redundancy evaluation and anti-interference capability based on the conditional probability calculation result.

[0154] The simulation report can not only support the decision-making of the current task, ensure the effective use and dynamic adjustment of simulation combat simulation resources, but also provide data basis for future task planning, ensure the reasonable allocation and efficient use of simulation combat simulation resources. Through the accumulation of evaluation results, the simulation combat simulation system based on knowledge graph and Bayesian network is continuously optimized and improved to improve the accuracy and practicality of the evaluation results.

[0155] The following is a detailed description with reference to specific embodiments:

[0156] Example 1: Figure 3 As shown, the simulated combat resource data is collected by ground radar sensors, dynamic reasoning of simulated combat simulation is performed, and corresponding simulation reports are obtained. The data collector includes multiple ground radar sensors.

[0157] The ground radar sensor collects the detection data of the simulated combat resources in real time. The collected simulated combat resource data includes the radar's detection range, target tracking data, radar working status, and the control instructions and feedback information sent by the command and control center to each ground radar sensor. Through a high-frequency data transmission link, the simulated combat resource data is transmitted to the command and control center in real time to ensure the timeliness and integrity of the data. The command and control center collects control instructions and feedback data.

[0158] In addition, the environmental monitoring sensor will collect the surrounding environment data of the ground radar sensor, such as weather conditions, terrain information, etc., and the maintenance station will collect equipment maintenance records and status data, as shown in Table 1. These data are of great significance for the evaluation of radar detection capabilities. In terms of simulation data, high-fidelity simulation software is used to simulate the actual situation of radar detection in a multi-target environment, and the generated simulation data is also used as an important input for the evaluation. The historical record data includes radar detection and control data in previous combat exercises and actual operations. These data provide rich background information and comparison basis for the evaluation.

[0159] Table 1 Data collection example table

[0160]

[0161] The collected simulated combat resource data is cleaned to remove noise and abnormal data caused by data collector failure or external interference to ensure the accuracy of the data; then format conversion is performed to convert the simulated combat resource data from different sources into a unified format to facilitate system processing; finally, normalization is performed to normalize the simulated combat resource data to the same dimension and range to ensure the comparability between different data.

[0162] The processed normalized data includes cleaned radar detection data, target tracking data, and command and control data. As shown in Table 2, after format conversion and normalization, these data ensure consistency between different data sources, which is convenient for subsequent analysis and processing.

[0163] Table 2 Data processing example table

[0164]

[0165] Entities are extracted from the normalized data, including radar sensor A, radar sensor B, radar sensor C (spare), command and control center 1, command and control center 2 (spare), communication link 1, communication link 2 (spare), environmental monitoring station 1, environmental monitoring station 2 (spare), maintenance station, etc. The nodes and node relationships in the structured graph of simulation combat resources are determined according to the normalized data. The attributes of each node are defined according to the normalized data, such as the technical parameters, functional characteristics, and geographical location of the radar sensor. Specifically, the radar sensor A and radar sensor B nodes record their detection and tracking data, the command and control center node records the control instructions and feedback information, and the communication link node records the data transmission between the radar sensor and the command center, as shown in Table 3.

[0166] Table 3 Entity attributes

[0167]

[0168] The weight of the edge is obtained through statistical analysis and correlation calculation, reflecting the synergy and interdependence between nodes. The relationship between each node is quantified by the data transmission success rate and command response time to form a complete structured map of simulation combat resources. By analyzing the data transmission and control relationship of each node, the importance and influence of each node are determined. Taking the data transmission relationship from radar sensor A to command and control center 1 as an example, it is assumed that the weight coefficients of each factor are as follows: α=0.9 (data transmission success rate weight), β=0.05 (command response time weight), γ=0.03 (environmental impact coefficient weight), δ=0.02 (redundancy level coefficient), the specific data from radar sensor A to command and control center 1: data transmission success rate S=99%, command response time T=50ms, environmental impact coefficient E=1.0, redundancy level coefficient R=1.0. Substituting the data into formula (3) yields:

[0169] W=0.9×99+0.05×(100−50)+0.03×1+0.02×1=89.1+2.5+0.03+0.02=91.65≈94.5.

[0170] In order to make the calculation results consistent with the weight table in the embodiment, the weight coefficient can be adjusted according to the specific system requirements and expert opinions. As shown in Table 4, some node relationships and their weight calculations are shown. The data transmission success rate and command response time directly affect the relationship weight. The environmental impact coefficient reflects the degree of influence of the environment on the relationship. The redundancy level coefficient reflects the impact of the redundant state of the entity on the relationship weight. The comprehensive weight is calculated using the weighted average method, and the weight coefficient is set according to the specific relationship type and system requirements. The nodes and node relationships are added to the graph database, and the key nodes and important relationships are identified through the graph theory algorithm, and the redundant relationships are eliminated. The final structured graph of simulated combat resources is as shown below. Figure 4 shown.

[0171] Table 4 Example of relationship weight calculation

[0172]

[0173] The Bayesian network model is optimized according to the established structured graph of simulated combat resources, and the Bayesian network model includes nodes: radar sensor, command and control center, communication link, environmental monitoring station, and maintenance station. First, the nodes of the Bayesian network model are initialized according to the node attributes in the structured graph of simulated combat resources, as shown in Table 5.

[0174] Table 5 Node mapping table

[0175]

[0176] A conditional probability table is constructed for each node in the Bayesian network to reflect the probability distribution of the node under different conditions. The conditional probability table is constructed by statistically analyzing indicators such as radar detection success rate, target tracking accuracy, and data transmission success rate to ensure that the probability distribution of each node under different conditions is accurate, as shown in Table 6.

[0177] Table 6 Conditional probability table

[0178]

[0179] The Bayesian network structure is optimized according to the node relationship in the structured graph of the simulated combat resources, which reduces redundant connections and improves computational efficiency and accuracy. The network reasoning efficiency and accuracy are ensured. Figure 5 As shown, reasoning and calculation can be performed more efficiently, providing more accurate performance evaluation results.

[0180] The accuracy and effectiveness of the Bayesian network model were verified through cross-tests, and the evaluation results were evaluated, as shown in Table 7.

[0181] Table 7 Verification results

[0182]

[0183] The impact of each node in the Bayesian network model on the overall performance was analyzed, and key nodes and weak links were identified, as shown in Table 8. According to the verification results, the structure and conditional probability table of the Bayesian network were adjusted to continuously optimize the model performance.

[0184] Table 8 Sensitivity analysis results

[0185]

[0186] The simulated combat resource data to be predicted and simulated is input into the optimized Bayesian network model, which performs dynamic reasoning and result evaluation based on the input simulated combat resource data and generates a detailed simulation report. The simulation report lists in detail the key indicators such as the detection success rate, target tracking accuracy and data transmission success rate of the radar sensor as a data collector under different mission conditions, providing comprehensive data support and improvement suggestions for command decision-making.

[0187] First, the simulation combat resource data is input into the Bayesian network model, including the operating status of each radar sensor (normal, under maintenance, faulty); the status of the communication link (normal, faulty); the environmental conditions of the environmental monitoring station (good, medium, basic); and the status of the maintenance station (available, unavailable), as shown in Table 9.

[0188] Table 9 Simulation combat resource data input table

[0189]

[0190] The Bayesian network model is used to calculate the conditional probability of each node and determine the posterior probability of each node.

[0191] Taking radar sensor A (RadarA_Status) and communication link 1 (CommunicationLink1_Status) as examples, calculate the detection success rate of radar sensor A. Assume that the status of radar sensor A is "normal" (RadarA_Status=normal); the status of communication link 1 is "normal" (CommunicationLink1_Status=normal); the environmental condition of environmental monitoring station 1 is "good" (Environment1_Status=good); according to the conditional probability table, calculate the detection success rate of radar sensor A under these conditions.

[0192] Calculate the detection success rate of RadarA: P(Detection_Success=Yes|RadarA_Status=Normal, CommunicationLink1_Status=Normal, Environment1_Status=Good)=0.98; P(Detection_Success=No|RadarA_Status=Normal, CommunicationLink1_Status=Normal, Environment1_Status=Good)=0.02.

[0193] RadarB is in a faulty state. Calculate the detection success rate of RadarB: P(Detection_Success=Yes|RadarB_Status=Faulty)=0.85; P(Detection_Success=No|RadarB_Status=Faulty)=0.15.

[0194] Calculate the transmission success rate of CommunicationLink1: P(CommunicationLink1_Status=Normal)=0.99.

[0195] According to the conditional probability calculation results, a simulation report including key indicators such as detection success rate, information transmission success rate, system redundancy evaluation and anti-interference capability is generated. As shown in Table 10, the simulation report lists the detection capability of each radar sensor under different conditions, the transmission efficiency of the communication link, the response efficiency of the command and control center, and the system redundancy and anti-interference capability.

[0196] Table 10 Simulation report example table

[0197]

[0198] The results show that the radar sensor as a data collector has a detection success rate of more than 80% in a multi-target environment and a target tracking accuracy of more than 90%. The report lists in detail the detection capabilities and target tracking accuracy of each radar sensor under different conditions, as well as the data transmission success rate and command response between the command and control center and the radar sensor.

[0199] The simulation report provides a scientific basis for command decision-making, points out which factors affect detection effectiveness, and puts forward suggestions for improvement. By analyzing the data of each radar sensor, it is possible to determine which radar sensors perform best in a specific environment, thereby optimizing the deployment and operation strategy of radar sensors. At the same time, the simulation report also points out potential problems in data transmission and command response of the command and control center, and provides improvement suggestions to improve the effectiveness of the overall system.

[0200] According to the evaluation results in the simulation report, the deployment of radar sensors in the current mission can be adjusted to improve the success rate of the mission. According to the data in the report, the command center can adjust the position and detection angle of the radar sensor in real time to optimize target tracking and information transmission. At the same time, the evaluation results provide data support for future radar sensor construction and upgrade plans, helping decision makers determine the best radar sensor deployment and resource allocation strategy. By optimizing the deployment of radar sensors, the detection and tracking efficiency can be improved, blind spots and coverage dead angles can be reduced, and the success rate and effectiveness of the mission can be enhanced. In addition, the evaluation results can also be used to formulate long-term radar sensor construction and upgrade plans to ensure the rational allocation and efficient use of resources and enhance the overall reconnaissance capability. In practical applications, the evaluation results provide important decision support for the current mission. The command center can adjust the position and detection angle of the radar sensor in real time according to the suggestions in the simulation report, optimize target tracking and information transmission, and improve the success rate and effectiveness of the mission. In terms of future mission planning, the evaluation results provide a scientific basis for the construction and upgrade of radar sensors, helping decision makers to formulate reasonable resource allocation and development plans, and ensure the continuous improvement and enhancement of the reconnaissance system.

[0201] Example 2: Figure 6As shown in the figure, the simulated combat resource data is collected by using the UAV sensor as a data collector, including multiple UAV sensors, ground control station and data transmission link. The UAV sensor collects flight data and detection data in real time, the ground control station monitoring equipment collects control instructions and feedback data, and the data recording system stores all data. The collected data includes the flight trajectory of the UAV, detection data, images and signal data captured by the sensor, as well as the control instructions sent by the ground control station and the feedback information received. These data are transmitted to the ground control station through a wireless transmission link for storage and processing, ensuring the real-time and integrity of the data. The ground control station also collects environmental data around the UAV sensor, such as weather conditions, terrain information, etc. These data are of great significance for the evaluation of the flight and detection capabilities of the UAV.

[0202] The simulated combat resource data collected by the UAV are sequentially subjected to data cleaning, format conversion, normalization and other operations to obtain normalized data. The processed normalized data includes cleaned UAV flight data, detection data and control data.

[0203] Establish a structured graph of simulation combat resources based on normalized data: First, determine the nodes and node attributes, including nodes: UAV A, UAV B, ground control station, data link and the transmission relationship between them. Specifically, the nodes of UAV A and UAV B record their flight and detection data, the ground control station node records the control instructions and feedback information, and the data link node records the data transmission between the UAV and the ground control station. The relationship between each node is quantified by the data transmission success rate and the command response time to form a complete knowledge graph.

[0204] Define attributes for the determined nodes, such as the technical parameters of the drone, flight capabilities, sensor performance, etc. Specifically, the drone A and drone B nodes record their flight and detection data, the ground control station node records control instructions and feedback information, and the data link node records the data transmission between the drone and the ground control station.

[0205] According to the quantification of the data transmission success rate and command response time between each node, the relationship between nodes is determined to form a structured map of simulation combat resources. By analyzing the data transmission and control relationship of each node, the importance and influence of each node in the system are determined.

[0206] First, data enhancement is performed, using deep learning technology to generate more training data and expand the size of the data set. Then, a deep learning model is used to extract key features from the training data, including drone flight mode, target detection mode, etc. By repeatedly training and optimizing the deep learning model, the prediction ability of drone sensors is improved.

[0207] The deep learning model includes but is not limited to a convolutional neural network.

[0208] The nodes in the Bayesian network model are initialized according to the nodes and node attributes of the structured graph of simulated combat resources. The nodes in the Bayesian network model include drone A, drone B, ground control station, data link, and conditional probability table. The conditional probability table is constructed by statistically analyzing indicators such as drone flight success rate, detection coverage, and data transmission success rate to ensure that the probability distribution of each node under different conditions is accurate.

[0209] The edges in the Bayesian network model are optimized according to the node relationships of the structured graph of simulated combat resources, thereby reducing redundant connections, improving computational efficiency and accuracy, and ensuring network reasoning efficiency and accuracy.

[0210] The simulated combat resource data to be predicted and simulated is input into the optimized Bayesian network model, which performs dynamic reasoning and result evaluation based on the input simulated combat resource data and generates a detailed simulation report. The simulation report lists in detail the key indicators such as the detection success rate, detection coverage and data transmission success rate of the UAV as a data collector under different mission conditions, providing comprehensive data support and improvement suggestions for command decision-making.

[0211] The results show that the detection success rate of drones as data collectors in complex environments is 87%, and the detection coverage reaches 92%. The report lists in detail the detection capabilities and coverage of each drone under different flight conditions, as well as the data transmission success rate and command response between the ground control station and the drone. The evaluation results provide a scientific basis for command decision-making, point out which factors affect the detection efficiency, and put forward suggestions for improvement. By analyzing the data of each drone, it is possible to determine which drones perform best under specific flight conditions, thereby optimizing the flight path and task allocation strategy of the drone. At the same time, the report also points out potential problems in data transmission and command response of the ground control station, and provides improvement suggestions to improve the effectiveness of the overall system.

[0212] The evaluation results are used to adjust the UAV flight path in the current mission and improve the mission success rate. According to the data in the simulation report, the command center can adjust the UAV flight path and task allocation in real time to optimize the detection angle and range. At the same time, the evaluation results provide data support for future UAV missions and deployment plans, helping decision makers determine the best UAV deployment and resource allocation strategy. By optimizing the flight path of the UAV, the detection efficiency can be improved, blind spots and coverage dead spots can be reduced, and the success rate and effectiveness of the mission can be enhanced. In addition, the evaluation results can also be used to formulate long-term UAV missions and deployment plans to ensure the rational allocation and efficient use of resources and enhance the overall reconnaissance capabilities. In practical applications, the evaluation results provide important decision-making support for the current mission. The command center can adjust the UAV flight path and task allocation in real time according to the suggestions in the simulation report, optimize the detection angle and range, and improve the success rate and effectiveness of the mission. In terms of future mission planning, the evaluation results provide a scientific basis for the mission and deployment of UAVs, helping decision makers to formulate reasonable resource allocation and development plans, and ensure the continuous improvement and enhancement of the reconnaissance system.

[0213] The combat simulation system and method based on knowledge graph and Bayesian network of the present invention can comprehensively understand the detection capability, synergy efficiency and overall detection effect of each data collector under different conditions, conduct combat simulation and generate corresponding simulation reports, thereby providing a scientific basis for command decision-making.

[0214] The above content is an explanation of the preferred embodiments of the present invention, which can help those skilled in the art to more fully understand the technical solution of the present invention. However, these embodiments are merely illustrative, and it cannot be determined that the specific implementation methods of the present invention are limited to the description of these embodiments. For ordinary technicians in the technical field to which the present invention belongs, without departing from the concept of the present invention, several simple deductions and transformations can be made, which should be regarded as belonging to the protection scope of the present invention.

Claims

1. A combat simulation system based on knowledge graph and Bayesian network, characterized in that: It includes a data collection unit, a data processing unit, a knowledge graph construction unit, a Bayesian network construction unit and a result evaluation unit; the data collection unit includes a plurality of data collectors, and the relevant data of the simulation combat resources are collected through the plurality of data collectors to obtain the simulation combat resource data; The data processing unit processes the collected simulated combat resource data to obtain normalized data; the data processing includes data cleaning, format conversion and normalization processing; The knowledge graph construction unit constructs a knowledge graph according to the normalized data to generate a structured graph of simulated combat resources; the knowledge graph construction unit includes a node determination module and a relationship determination module, the node determination module identifies and defines the data collector and its attributes, and the relationship determination module analyzes the synergy between the various data collectors to determine the relationship between the nodes; The Bayesian network construction unit obtains a Bayesian network model according to the nodes and node relationships of the simulated combat resource structured graph; initializes the nodes of the Bayesian network model according to the simulated combat resource structured graph; then constructs a conditional probability table corresponding to each node in the Bayesian network model based on statistical analysis and actual data; optimizes the structure of the Bayesian network model, verifies the Bayesian network model, and evaluates the performance indicators of the Bayesian network model; Analyze the impact of each node in the Bayesian network on the overall performance, identify key nodes and weak links, and evaluate the marginal influence of each node on the target node: ， Among them, Impact(X i ) represents node X i The degree of influence on the target node; States(X i ) represents node X i All possible states of P(Target|X i =x) means the target node is at X i The probability of being in state x; By calculating the influence of the conditional probability changes between nodes on the probability distribution of the target node, the structure and conditional probability table of the Bayesian network model are adjusted to continuously optimize the performance of the Bayesian network model; The result evaluation unit performs dynamic reasoning of the simulated combat simulation according to the Bayesian network model and generates a simulation report.

2. The combat simulation system based on knowledge graph and Bayesian network according to claim 1 is characterized in that: The data cleaning includes calculating the mean μ and standard deviation σ of the data points in the data set. ， ， Where N is the total number of data points in the dataset, X i represents the i-th data point; Then calculate the Z-Score value of each data point. ， Finally, an abnormal threshold is set, and data points whose absolute value of Z-Score is greater than the abnormal threshold are removed.

3. The combat simulation system based on knowledge graph and Bayesian network according to claim 1, characterized in that: The normalization process adopts Min-Max normalization or Z-Score standardization technology. When the type of the simulated combat resource data is in JSON format, Min-Max normalization is selected; when the type of the simulated combat resource data is in CSV format, Z-Score standardization is selected to scale the simulated combat resource data to the same dimension and range. , ， Where X represents the data point; X min Indicates the minimum value among the data points; X max represents the maximum value of the data point; μ represents the mean of the data set; σ represents the standard deviation of the data set.

4. The combat simulation system based on knowledge graph and Bayesian network according to claim 1 is characterized in that: Each node represents a data collector, and the properties of each node are defined according to the data collector corresponding to the node.

5. The combat simulation system based on knowledge graph and Bayesian network according to claim 1 is characterized in that: The relationship determination module determines the direction and weight of the edge by calculating the comprehensive weight W. The directed edge represents the direction of the relationship between nodes, and the weight represents the strength of the relationship, including the data transmission success rate S, the command response time T, the environmental impact coefficient E and the redundancy level coefficient R. ， Among them, α, β, γ, and δ represent the weight coefficients of each factor, respectively, satisfying α+β+γ+δ=1.

6. The combat simulation system based on knowledge graph and Bayesian network according to claim 1 is characterized in that: Each node has a conditional probability table, which represents the probability distribution of the node under different conditions. For nodes without parent nodes, the prior probability is calculated directly: ， Among them, N(X i =x) represents node X i The number of nodes in state x; N represents the number of nodes X i Total number of For nodes with parent nodes, the conditional probability is calculated based on the state of its parent node and the actual data: ， Among them, X i Represents the current node; x represents a state of the current node; p represents the state combination of the parent node.

7. The combat simulation system based on knowledge graph and Bayesian network according to claim 1 is characterized in that: Use cross-validation to verify the Bayesian network model; divide the data set into a training set and a test set, use the training set to train the Bayesian network model, and evaluate the performance indicators of the Bayesian network model on different data sets on the test set, adjust the structure and conditional probability table of the Bayesian network, and continuously optimize the Bayesian network model.

8. The combat simulation system based on knowledge graph and Bayesian network according to claim 7 is characterized in that: The performance evaluation includes precision, recall and F1 score. Evaluate the accuracy of the Bayesian network model: ， Among them, TP represents the number of positive classes that are actually predicted to be positive; TN represents the number of negative classes that are actually predicted to be negative; FP represents the number of negative classes that are actually predicted to be positive; FN represents the number of positive classes that are actually predicted to be negative. Evaluate the recall of the Bayesian network model, that is, the proportion of correct predictions in actual positive examples: ， Among them, TP represents the number of positive classes that are actually predicted to be positive; FN represents the number of negative classes that are actually predicted to be positive; Evaluate the F1 score of the Bayesian network model, which is the harmonic mean of precision and recall: 。 9. A combat simulation method based on knowledge graph and Bayesian network, applicable to the combat simulation system based on knowledge graph and Bayesian network, characterized in that: The specific steps include: Step 1: The data acquisition unit collects relevant data of the simulated combat resources through a variety of data collectors to obtain simulated combat resource data; Step 2: The data processing unit processes the collected simulation combat resource data to obtain normalized data; the data processing includes data cleaning, format conversion and normalization processing; Step 3: The knowledge graph construction unit constructs a knowledge graph according to the normalized data to generate a structured graph of simulated combat resources; the knowledge graph construction unit includes a node determination module and a relationship determination module, the node determination module identifies and defines the data collector and its attributes, and the relationship determination module analyzes the synergy between the data collectors and determines the relationship between the nodes; Step 4, the Bayesian network construction unit obtains a Bayesian network model according to the nodes and node relationships of the simulated combat resource structured graph; initializes the nodes of the Bayesian network model according to the simulated combat resource structured graph; then constructs a conditional probability table corresponding to each node in the Bayesian network model based on statistical analysis and actual data; optimizes the structure of the Bayesian network model, verifies the Bayesian network model, and evaluates the performance indicators of the Bayesian network model; Analyze the impact of each node in the Bayesian network on the overall performance, identify key nodes and weak links, and evaluate the marginal influence of each node on the target node: ， Among them, Impact(X i ) represents node X i The degree of influence on the target node; States(X i ) represents node X i All possible states of P(Target|X i =x) means the target node is at X i The probability of being in state x; By calculating the influence of the conditional probability changes between nodes on the probability distribution of the target node, the structure and conditional probability table of the Bayesian network model are adjusted to continuously optimize the performance of the Bayesian network model; Step 5: The result evaluation unit performs dynamic reasoning of the simulated combat simulation based on the Bayesian network model and generates a simulation report.

Images