Simulation verification result intelligent diagnosis method and system based on knowledge graph

Abstract

The invention discloses a simulation verification result intelligent diagnosis method and system based on a knowledge graph, and relates to the cross technical field of a simulation verification technology, an intelligent diagnosis technology and a knowledge graph technology. The method comprises the following steps: firstly, constructing a domain knowledge graph containing simulation object attributes, verification standard thresholds, fault modes and diagnosis rules; obtaining simulation verification data to be diagnosed and extracting time sequence / non-time sequence features; associating the feature data with the knowledge graph through a hierarchical matching strategy, and screening abnormal features; a mixed mechanism of rule reasoning and graph neural network reasoning is adopted to determine a preliminary fault mode; and outputting a final diagnosis result and a solution through secondary verification, and establishing a knowledge graph updating and diagnosis result feedback mechanism at the same time. The problems that traditional simulation diagnosis depends on manpower, efficiency is low and knowledge is difficult to reuse are solved, fault positioning accuracy and diagnosis efficiency are remarkably improved, and the method can adapt to a distributed simulation scene and has good expansibility and timeliness.

Description

Technical Field

[0001] This invention relates to the interdisciplinary field of simulation verification technology, intelligent diagnosis technology and knowledge graph technology, specifically to an intelligent diagnosis method and system for simulation verification results based on knowledge graphs. Background Technology

[0002] In the development of complex systems, simulation verification is a crucial step in verifying the rationality of system design and identifying potential faults. As system complexity increases, the amount of data generated by simulation verification grows exponentially, involving multiple types of performance parameters and multiple simulation scenarios.

[0003] Traditional simulation verification result diagnosis methods have the following significant shortcomings: Existing diagnoses largely rely on engineers manually comparing simulation data with standard thresholds and analyzing the correlation between abnormal parameters and faults. For scenarios with multiple parameters exhibiting concurrent anomalies, this requires a significant amount of time for troubleshooting, making it difficult to meet real-time diagnostic needs. Engineers' diagnostic experience is mostly stored in document form, lacking structured organization. New engineers need long-term learning to master this knowledge, and fault diagnosis knowledge is difficult to share between different projects. Traditional methods lack in-depth analysis of parameter correlations, making it easy to misjudge the root cause of the fault when multiple parameters are abnormal. For distributed simulation systems, existing data acquisition and diagnostic methods struggle to achieve cross-node data integration and lack data security protection mechanisms. When industry standards iterate and new fault modes emerge, the existing diagnostic system struggles to quickly incorporate new knowledge, leading to a disconnect between diagnostic results and actual needs.

[0004] Current diagnostic methods largely rely on engineers manually comparing simulation data with standard thresholds, which consumes a significant amount of time and is insufficient for real-time diagnostic needs. Engineers' diagnostic experience is mostly stored in document form, requiring new engineers to undergo extensive learning to master it, and fault diagnosis knowledge is difficult to share. Traditional methods lack in-depth analysis of parameter correlations, making it easy to misjudge the root cause of a fault when multiple parameters are abnormal. Furthermore, for distributed simulation systems, there is a lack of data security protection mechanisms. When industry standards iterate and new fault modes emerge, the existing diagnostic system struggles to quickly incorporate new knowledge, leading to a disconnect between diagnostic results and actual needs. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent diagnostic method and system for simulation verification results based on knowledge graphs, in order to solve the technical problems of "low efficiency, difficulty in knowledge reuse, inaccurate positioning, poor adaptability, and lagging updates" in the existing simulation verification result diagnosis mentioned in the background art, and to realize the automation, intelligence, and scalability of simulation diagnosis, while ensuring data security and real-time knowledge updates in distributed scenarios.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent diagnostic method and system for simulation verification results based on knowledge graphs, comprising the following steps: S1. Construct a knowledge graph for simulation verification, wherein the knowledge graph includes simulation object attribute information, simulation verification standard threshold information, simulation fault mode information and fault diagnosis rule information. The simulation object attribute information includes object structure parameters and performance parameters. The simulation verification standard threshold information is the qualified range of each performance parameter under different simulation scenarios. The simulation fault mode information corresponds to the fault type when different performance parameters are abnormal. The fault diagnosis rule information is the association logic between abnormal performance parameters and fault modes. S2. Obtain the simulation verification result data to be diagnosed, wherein the simulation verification result data includes the time series data of each performance parameter collected during the simulation process, the simulation scenario parameters, and the simulation object identification information; S3. Perform feature extraction on the simulation verification result data to obtain feature data. If the simulation verification result data is time series data, extract time series features including data mean, variance, peak value, valley value and trend change rate. If it is non-time series data, extract attribute features including the deviation between the actual value of the parameter and the preset benchmark value. S4. Match the feature data with the simulation verification domain knowledge graph constructed in step S1. First, locate the corresponding simulation object node in the knowledge graph based on the simulation object identification information. Then, compare the extracted feature data with the simulation verification standard threshold information associated with the node to filter out abnormal feature data that exceed the standard threshold range. S5. Based on the abnormal feature data and the fault diagnosis rule information in the knowledge graph, reasoning is performed to determine the corresponding simulation fault mode. If there are multiple abnormal feature data during the reasoning process, the matching degree of each fault mode is calculated by the association weight between the fault mode and multiple abnormal features in the knowledge graph, and the fault mode with the highest matching degree is selected as the preliminary diagnosis result. S6. Verify the preliminary diagnosis result by calling the typical feature data corresponding to the fault mode in the knowledge graph and comparing it with the abnormal feature data of the simulation verification result to be diagnosed. If the feature matching degree of the two is greater than the preset threshold of 85%, the fault mode is output as the final diagnosis result. If it is less than or equal to 85%, return to step S3 to re-optimize the feature extraction dimension and then execute the subsequent steps again.

[0007] Preferably, the specific process of constructing the simulation verification domain knowledge graph in step S1 includes: S11. Collect data in the field of simulation verification. Data sources include simulation object design documents, historical simulation verification reports, industry standards and specifications, and expert experience data. Among them, historical simulation verification reports include fault cases and corresponding diagnostic results from past simulations. S12. Preprocess the collected data, including data cleaning to remove redundant and erroneous data, data standardization to unify parameter units and formats, and data structuring to convert unstructured text data into key-value pair format; S13. Based on the preprocessed data, perform ontology modeling, define the core concept classes of the knowledge graph, including "simulation object", "performance parameter", "verification standard", "fault mode" and "diagnostic rule", and determine the inter-class relationships, including "object-contains-parameter", "parameter-correspondence-standard", "parameter anomaly-trigger-fault", and "fault-follows-rule". S14. Use a graph database to store the ontology model and associated data, and construct the nodes and edges of the knowledge graph. The node attributes include the specific attribute values ​​of the class, and the edge attributes include the weight and confidence of the association. At the same time, establish an indexing mechanism for the knowledge graph to quickly locate the simulation object nodes and associated information.

[0008] Preferably, when obtaining the simulation verification result data to be diagnosed in step S2, if the simulation system is a distributed simulation system, data is obtained from each simulation sub-node in real time through a distributed data acquisition protocol. During data transmission, an encryption algorithm is used to protect data security. The encryption algorithm is AES-256. At the same time, a timestamp and data source identifier are added to the acquired data to trace the data acquisition node and acquisition time.

[0009] Preferably, the feature extraction in step S3 adopts a fusion algorithm. For time-series performance parameter data, wavelet transform is first used to remove data noise, and then deep time-series features are extracted through LSTM network and fused with traditional statistical features to form the final time-series feature data. For non-time-series data, fuzzy mathematics is used to fuzzify the parameter deviation to obtain fuzzy feature vectors as attribute feature data.

[0010] Preferably, in step S4, the matching of feature data and knowledge graph adopts a hierarchical matching strategy. The first layer is attribute matching, which accurately matches the simulation object identification information with the attributes of the "simulation object" class node in the knowledge graph to locate the target node. The second layer is threshold matching, which performs interval matching between the actual value of the parameter in the extracted feature data and the threshold range of the "verification standard" class node associated with the target node, and marks abnormal parameters that exceed the interval. The third layer is semantic matching, which semantically associates the abnormal parameters with the "performance parameter" class node through the semantic relationships in the knowledge graph to ensure the accuracy of the matching.

[0011] Preferably, the reasoning process in step S5 adopts a hybrid reasoning mechanism, combining rule-based reasoning and graph neural network-based reasoning. Rule-based reasoning utilizes the association logic stored in the "diagnosis rule" type nodes in the knowledge graph to initially screen fault modes. Graph neural network-based reasoning uses the knowledge graph as the topology structure, takes abnormal feature data as node input features, learns the association weights between nodes through the GCN network, calculates the score of each fault mode node, and the fault mode with the highest score is the preliminary diagnosis result. The association weights are obtained by training with historical fault case data.

[0012] Preferably, the verification process in step S6 also includes calling the "solution" information corresponding to the fault mode in the knowledge graph. If the final diagnosis result is determined, the corresponding fault solution is output synchronously. The solution includes fault troubleshooting steps, parameter adjustment suggestions, and simulation system optimization directions. If the verification fails and step S3 is re-executed, the feature extraction index of the unmatched feature dimension is added, such as adding the periodic feature of time series data or the deviation change rate feature of non-time series data.

[0013] Preferably, in step S14, the knowledge graph is also equipped with an update mechanism, which regularly collects new simulation verification data, fault cases and industry standard update data, and updates the nodes and edges of the knowledge graph through incremental learning algorithm. During the update process, the new data is preprocessed first, and then conflict detection is performed with the existing knowledge graph. If there is a data conflict, the expert review result shall be used to determine whether to update, so as to ensure the timeliness and accuracy of the knowledge graph.

[0014] Preferably, the intelligent diagnosis method for simulation verification results based on knowledge graphs further includes step S7: establishing a diagnosis result feedback mechanism, storing the final result of each diagnosis and the corresponding simulation verification data and feature data in a historical database, periodically analyzing the data in the historical database, mining new fault modes and diagnosis rules, adding the new fault modes as new nodes in the "fault mode" class and the new diagnosis rules as new nodes in the "diagnosis rule" class to the knowledge graph, thereby realizing the self-improvement of the knowledge graph.

[0015] Preferably, a knowledge graph-based intelligent diagnostic system for simulation verification results includes: The knowledge graph construction module is used to collect data from the simulation verification domain, perform data preprocessing, ontology modeling, and graph database storage, and construct and update the knowledge graph of the simulation verification domain. The module includes a data acquisition unit, a data preprocessing unit, an ontology modeling unit, a graph storage unit, and a graph update unit. The data acquisition unit is used to collect data such as design documents and historical reports. The data preprocessing unit performs data cleaning and standardization. The ontology modeling unit defines concept classes and relationships. The graph storage unit uses a graph database to store graph data. The graph update unit realizes incremental updates of the graph. The simulation data acquisition module is used to acquire the simulation verification result data to be diagnosed. It includes a distributed data acquisition unit, a data encryption unit, and a data identification unit. The distributed data acquisition unit collects data from the simulation node, the data encryption unit encrypts the data using the AES-256 algorithm, and the data identification unit adds a timestamp and source identifier. The feature extraction module is used to extract features from the simulation verification result data. It includes a temporal feature extraction unit, a non-temporal feature extraction unit, and a feature fusion unit. The temporal feature extraction unit uses wavelet transform and LSTM network to extract temporal features. The non-temporal feature extraction unit uses fuzzy mathematics to extract attribute features. The feature fusion unit fuses different types of features to obtain feature data. The graph matching module is used to match feature data with knowledge graphs. It includes attribute matching unit, threshold matching unit and semantic matching unit, which respectively realize object localization, abnormal parameter marking and semantic association matching. The diagnostic reasoning module is used to determine the fault mode based on abnormal feature data. It includes a rule reasoning unit, a graph neural network reasoning unit, and a matching degree calculation unit. The rule reasoning unit filters fault modes based on diagnostic rules, the graph neural network reasoning unit calculates the fault mode score through the GCN network, and the matching degree calculation unit determines the preliminary diagnostic result. The result verification and output module is used to verify the preliminary diagnosis results and output the final results and solutions. It includes a result verification unit, a solution generation unit, and a result feedback unit. The result verification unit performs secondary comparison and verification, the solution generation unit outputs the fault solution, and the result feedback unit feeds back the diagnosis data to the historical database. The historical database module is used to store historical diagnostic data. It includes a data storage unit and a data analysis unit. The data storage unit stores diagnostic results, simulation data, and feature data. The data analysis unit mines new fault modes and diagnostic rules, providing data support for the improvement of the knowledge graph.

[0016] The technical effects and advantages of this invention are as follows: 1. This intelligent diagnostic method and system for simulation verification results based on knowledge graphs automates data collection, feature extraction, and reasoning diagnosis, replacing manual operations and reducing diagnosis time by more than 60%. 2. The intelligent diagnostic method and system based on knowledge graph simulation verification results reduces parameter association misjudgment through hierarchical matching and hybrid reasoning mechanisms, and improves the fault location accuracy to over 90%. 3. This intelligent diagnostic method and system for simulation verification results based on knowledge graphs uses knowledge graphs to store diagnostic knowledge in a structured manner, which can be directly accessed across projects and engineers, reducing learning costs. 4. This intelligent diagnostic method and system for simulation verification results based on knowledge graphs supports distributed data acquisition and encrypted transmission, meeting the diagnostic needs of complex simulation systems. 5. The intelligent diagnostic method and system for simulation verification results based on knowledge graphs, with incremental update and feedback mechanisms, ensure that the knowledge graph dynamically iterates with industry standards and new fault modes, avoiding diagnostic lag. 6. This intelligent diagnostic method and system for simulation verification results based on knowledge graphs synchronously outputs fault solutions, helping engineers quickly troubleshoot problems and optimize simulation system design. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the overall system flow of the present invention; Figure 2 This is a flowchart illustrating the knowledge graph construction module of the present invention; Figure 3 This is a flowchart illustrating the simulation data acquisition module of the present invention. Figure 4 This is a flowchart illustrating the feature extraction module and the map matching module of the present invention; Figure 5 This is a flowchart illustrating the diagnostic reasoning module, result verification and output module, and historical database module of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] This invention discloses an intelligent diagnostic method and system for simulation verification results based on knowledge graphs, according to the appendix. Figure 1 As shown, it includes the following steps: S1. Construct a knowledge graph for simulation verification, wherein the knowledge graph includes simulation object attribute information, simulation verification standard threshold information, simulation fault mode information and fault diagnosis rule information. The simulation object attribute information includes object structure parameters and performance parameters. The simulation verification standard threshold information is the qualified range of each performance parameter under different simulation scenarios. The simulation fault mode information corresponds to the fault type when different performance parameters are abnormal. The fault diagnosis rule information is the association logic between abnormal performance parameters and fault modes. S2. Obtain the simulation verification result data to be diagnosed, wherein the simulation verification result data includes the time series data of each performance parameter collected during the simulation process, the simulation scenario parameters, and the simulation object identification information; S3. Perform feature extraction on the simulation verification result data to obtain feature data. If the simulation verification result data is time series data, extract time series features including data mean, variance, peak value, valley value and trend change rate. If it is non-time series data, extract attribute features including the deviation between the actual value of the parameter and the preset benchmark value. S4. Match the feature data with the simulation verification domain knowledge graph constructed in step S1. First, locate the corresponding simulation object node in the knowledge graph based on the simulation object identification information. Then, compare the extracted feature data with the simulation verification standard threshold information associated with the node to filter out abnormal feature data that exceed the standard threshold range. S5. Based on the abnormal feature data and the fault diagnosis rule information in the knowledge graph, reasoning is performed to determine the corresponding simulation fault mode. If there are multiple abnormal feature data during the reasoning process, the matching degree of each fault mode is calculated by the association weight between the fault mode and multiple abnormal features in the knowledge graph, and the fault mode with the highest matching degree is selected as the preliminary diagnosis result. S6. Verify the preliminary diagnosis result by calling the typical feature data corresponding to the fault mode in the knowledge graph and comparing it with the abnormal feature data of the simulation verification result to be diagnosed. If the feature matching degree of the two is greater than the preset threshold of 85%, the fault mode is output as the final diagnosis result. If it is less than or equal to 85%, return to step S3 to re-optimize the feature extraction dimension and then execute the subsequent steps again.

[0021] According to the appendix Figures 1-2 As shown, further, the specific process of constructing the simulation verification domain knowledge graph in step S1 includes: S11. Collect data in the field of simulation verification. Data sources include simulation object design documents, historical simulation verification reports, industry standards and specifications, and expert experience data. Among them, historical simulation verification reports include fault cases and corresponding diagnostic results from past simulations. S12. Preprocess the collected data, including data cleaning to remove redundant and erroneous data, data standardization to unify parameter units and formats, and data structuring to convert unstructured text data into key-value pair format; S13. Based on the preprocessed data, perform ontology modeling, define the core concept classes of the knowledge graph, including "simulation object", "performance parameter", "verification standard", "fault mode" and "diagnostic rule", and determine the inter-class relationships, including "object-contains-parameter", "parameter-correspondence-standard", "parameter anomaly-trigger-fault", and "fault-follows-rule". S14. Use a graph database to store the ontology model and associated data, and construct the nodes and edges of the knowledge graph. The node attributes include the specific attribute values ​​of the class, and the edge attributes include the weight and confidence of the association. At the same time, establish an indexing mechanism for the knowledge graph to quickly locate the simulation object nodes and associated information.

[0022] According to the appendix Figures 1 to 3 As shown, further, in step S2, when obtaining the simulation verification result data to be diagnosed, if the simulation system is a distributed simulation system, data is obtained in real time from each simulation sub-node through a distributed data acquisition protocol. During data transmission, an encryption algorithm is used to protect data security. The encryption algorithm is AES-256. At the same time, a timestamp and data source identifier are added to the acquired data to trace the data acquisition node and acquisition time.

[0023] According to the appendix Figures 1 to 4 As shown, in particular, the feature extraction in step S3 employs a fusion algorithm. For time-series performance parameter data, wavelet transform is first used to remove data noise, and then deep time-series features are extracted through an LSTM network and fused with traditional statistical features to form the final time-series feature data. For non-time-series data, fuzzy mathematics is used to fuzzify the parameter deviation to obtain fuzzy feature vectors as attribute feature data.

[0024] According to the appendix Figures 1 to 4 As shown, specifically disclosed, the feature data and knowledge graph matching in step S4 adopts a hierarchical matching strategy. The first layer is attribute matching, which accurately matches the simulation object identification information with the attributes of the "simulation object" class node in the knowledge graph to locate the target node. The second layer is threshold matching, which performs interval matching between the actual values ​​of the parameters in the extracted feature data and the threshold range of the "verification standard" class node associated with the target node, and marks abnormal parameters that exceed the interval. The third layer is semantic matching, which semantically associates the abnormal parameters with the "performance parameter" class node through the semantic relationships in the knowledge graph to ensure the accuracy of the matching.

[0025] According to the appendix Figures 1 to 5As shown, it is particularly important to emphasize that the reasoning process in step S5 adopts a hybrid reasoning mechanism, combining rule-based reasoning and graph neural network-based reasoning. Rule-based reasoning utilizes the association logic stored in the "diagnostic rule" type nodes in the knowledge graph to initially screen fault modes. Graph neural network-based reasoning uses the knowledge graph as its topology, takes abnormal feature data as node input features, learns the association weights between nodes through the GCN network, calculates the score of each fault mode node, and the fault mode with the highest score is the preliminary diagnosis result. The association weights are obtained through training with historical fault case data.

[0026] According to the appendix Figures 1 to 5 As shown, it is particularly important to emphasize that the verification process in step S6 also includes calling the "solution" information corresponding to the fault mode in the knowledge graph. If the final diagnosis result is determined, the corresponding fault solution is output synchronously. The solution includes fault troubleshooting steps, parameter adjustment suggestions, and simulation system optimization directions. If the verification fails and step S3 is re-executed, feature extraction indicators for the unmatched feature dimensions are added, such as adding periodic features of time series data or deviation change rate features of non-time series data.

[0027] According to the appendix Figures 1 to 5 As shown, it is particularly important to emphasize that the knowledge graph in step S14 is also equipped with an update mechanism. New simulation verification data, fault cases and industry standard update data are collected regularly. The nodes and edges of the knowledge graph are updated through incremental learning algorithm. During the update process, the new data is preprocessed first, and then conflict detection is performed with the existing knowledge graph. If there is a data conflict, the expert review result shall be used to determine whether to update, so as to ensure the timeliness and accuracy of the knowledge graph.

[0028] According to the appendix Figures 1 to 5 As shown, it is particularly important to emphasize step S7: establish a diagnostic result feedback mechanism, store the final results of each diagnosis and the corresponding simulation verification data and feature data in the historical database, analyze the data in the historical database regularly, explore new fault modes and diagnostic rules, add the new fault modes as new nodes of the "fault mode" class and the new diagnostic rules as new nodes of the "diagnostic rule" class to the knowledge graph, so as to realize the self-improvement of the knowledge graph.

[0029] According to the appendix Figures 1 to 5 As shown, it is particularly important to emphasize that a knowledge graph-based intelligent diagnostic system for simulation verification results includes: The knowledge graph construction module is used to collect data from the simulation verification domain, perform data preprocessing, ontology modeling, and graph database storage, and construct and update the knowledge graph of the simulation verification domain. The module includes a data acquisition unit, a data preprocessing unit, an ontology modeling unit, a graph storage unit, and a graph update unit. The data acquisition unit is used to collect data such as design documents and historical reports. The data preprocessing unit performs data cleaning and standardization. The ontology modeling unit defines concept classes and relationships. The graph storage unit uses a graph database to store graph data. The graph update unit realizes incremental updates of the graph. The simulation data acquisition module is used to acquire the simulation verification result data to be diagnosed. It includes a distributed data acquisition unit, a data encryption unit, and a data identification unit. The distributed data acquisition unit collects data from the simulation node, the data encryption unit encrypts the data using the AES-256 algorithm, and the data identification unit adds a timestamp and source identifier. The feature extraction module is used to extract features from the simulation verification result data. It includes a temporal feature extraction unit, a non-temporal feature extraction unit, and a feature fusion unit. The temporal feature extraction unit uses wavelet transform and LSTM network to extract temporal features. The non-temporal feature extraction unit uses fuzzy mathematics to extract attribute features. The feature fusion unit fuses different types of features to obtain feature data. The graph matching module is used to match feature data with knowledge graphs. It includes attribute matching unit, threshold matching unit and semantic matching unit, which respectively realize object localization, abnormal parameter marking and semantic association matching. The diagnostic reasoning module is used to determine the fault mode based on abnormal feature data. It includes a rule reasoning unit, a graph neural network reasoning unit, and a matching degree calculation unit. The rule reasoning unit filters fault modes based on diagnostic rules, the graph neural network reasoning unit calculates the fault mode score through the GCN network, and the matching degree calculation unit determines the preliminary diagnostic result. The result verification and output module is used to verify the preliminary diagnosis results and output the final results and solutions. It includes a result verification unit, a solution generation unit, and a result feedback unit. The result verification unit performs secondary comparison and verification, the solution generation unit outputs the fault solution, and the result feedback unit feeds back the diagnosis data to the historical database. The historical database module is used to store historical diagnostic data. It includes a data storage unit and a data analysis unit. The data storage unit stores diagnostic results, simulation data, and feature data. The data analysis unit mines new fault modes and diagnostic rules, providing data support for the improvement of the knowledge graph.

[0030] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0031] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent diagnosis of simulation verification results based on knowledge graphs, characterized in that, Includes the following steps: S1. Construct a knowledge graph for simulation verification, wherein the knowledge graph includes simulation object attribute information, simulation verification standard threshold information, simulation fault mode information and fault diagnosis rule information. The simulation object attribute information includes object structure parameters and performance parameters. The simulation verification standard threshold information is the qualified range of each performance parameter under different simulation scenarios. The simulation fault mode information corresponds to the fault type when different performance parameters are abnormal. The fault diagnosis rule information is the association logic between abnormal performance parameters and fault modes. S2. Obtain the simulation verification result data to be diagnosed, wherein the simulation verification result data includes the time series data of each performance parameter collected during the simulation process, the simulation scenario parameters, and the simulation object identification information; S3. Perform feature extraction on the simulation verification result data to obtain feature data. If the simulation verification result data is time series data, extract time series features including data mean, variance, peak value, valley value and trend change rate. If it is non-time series data, extract attribute features including the deviation between the actual value of the parameter and the preset benchmark value. S4. Match the feature data with the simulation verification domain knowledge graph constructed in step S1. First, locate the corresponding simulation object node in the knowledge graph based on the simulation object identification information. Then, compare the extracted feature data with the simulation verification standard threshold information associated with the node to filter out abnormal feature data that exceed the standard threshold range. S5. Based on the abnormal feature data and the fault diagnosis rule information in the knowledge graph, reasoning is performed to determine the corresponding simulation fault mode. If there are multiple abnormal feature data during the reasoning process, the matching degree of each fault mode is calculated by the association weight between the fault mode and multiple abnormal features in the knowledge graph, and the fault mode with the highest matching degree is selected as the preliminary diagnosis result. S6. Verify the preliminary diagnosis result by calling the typical feature data corresponding to the fault mode in the knowledge graph and comparing it with the abnormal feature data of the simulation verification result to be diagnosed. If the feature matching degree of the two is greater than the preset threshold of 85%, the fault mode is output as the final diagnosis result. If it is less than or equal to 85%, return to step S3 to re-optimize the feature extraction dimension and then execute the subsequent steps again.

2. The intelligent diagnostic method for simulation verification results based on knowledge graphs according to claim 1, characterized in that, The specific process of constructing the simulation verification domain knowledge graph in step S1 includes: S11. Collect data in the field of simulation verification. Data sources include simulation object design documents, historical simulation verification reports, industry standards and specifications, and expert experience data. Among them, historical simulation verification reports include fault cases and corresponding diagnostic results from past simulations. S12. Preprocess the collected data, including data cleaning to remove redundant and erroneous data, data standardization to unify parameter units and formats, and data structuring to convert unstructured text data into key-value pair format; S13. Based on the preprocessed data, perform ontology modeling, define the core concept classes of the knowledge graph, including "simulation object", "performance parameter", "verification standard", "fault mode" and "diagnostic rule", and determine the inter-class relationships, including "object-contains-parameter", "parameter-correspondence-standard", "parameter anomaly-trigger-fault", and "fault-follows-rule". S14. Use a graph database to store the ontology model and associated data, and construct the nodes and edges of the knowledge graph. The node attributes include the specific attribute values ​​of the class, and the edge attributes include the weight and confidence of the association. At the same time, establish an indexing mechanism for the knowledge graph to quickly locate the simulation object nodes and associated information.

3. The intelligent diagnostic method for simulation verification results based on knowledge graphs according to claim 1, characterized in that, In step S2, when obtaining the simulation verification result data to be diagnosed, if the simulation system is a distributed simulation system, data is obtained in real time from each simulation sub-node through a distributed data acquisition protocol. During data transmission, an encryption algorithm is used to protect data security. The encryption algorithm is AES-256. At the same time, a timestamp and data source identifier are added to the acquired data to trace the data acquisition node and acquisition time.

4. The intelligent diagnostic method for simulation verification results based on knowledge graphs according to claim 1, characterized in that, In step S3, the feature extraction adopts a fusion algorithm. For time series performance parameter data, wavelet transform is first used to remove data noise, and then deep time series features are extracted through LSTM network and fused with traditional statistical features to form the final time series feature data. For non-time-series data, fuzzy mathematics is used to fuzzify the parameter deviations, resulting in fuzzy feature vectors as attribute feature data.

5. The intelligent diagnostic method for simulation verification results based on knowledge graphs according to claim 1, characterized in that, In step S4, the feature data and knowledge graph matching adopt a hierarchical matching strategy. The first layer is attribute matching, which accurately matches the simulation object identification information with the attributes of the "simulation object" class node in the knowledge graph to locate the target node. The second layer is threshold matching, which matches the actual values ​​of parameters in the extracted feature data with the threshold range of the "verification standard" class nodes associated with the target node, and marks abnormal parameters that exceed the range; the third layer is semantic matching, which semantically associates abnormal parameters with "performance parameter" class nodes through semantic relationships in the knowledge graph to ensure the accuracy of the matching.

6. The intelligent diagnostic method for simulation verification results based on knowledge graphs according to claim 1, characterized in that, In step S5, the reasoning process adopts a hybrid reasoning mechanism, combining rule-based reasoning and graph neural network-based reasoning. Rule-based reasoning utilizes the association logic stored in the "diagnostic rule" type nodes in the knowledge graph to initially screen fault modes. Graph neural network-based reasoning uses the knowledge graph as its topology, takes abnormal feature data as node input features, learns the association weights between nodes through the GCN network, calculates the score of each fault mode node, and the fault mode with the highest score is the preliminary diagnosis result. The association weights are obtained by training with historical fault case data.

7. The intelligent diagnostic method for simulation verification results based on knowledge graphs according to claim 1, characterized in that, The verification process in step S6 also includes calling the "solution" information corresponding to the fault mode in the knowledge graph. If the final diagnosis result is determined, the corresponding fault solution is output synchronously. The solution includes fault troubleshooting steps, parameter adjustment suggestions and simulation system optimization directions. If the verification fails and step S3 is re-executed, based on the unmatched feature dimension, the feature extraction index for that dimension is increased, such as adding the periodic feature of time series data or the deviation change rate feature of non-time series data.

8. The intelligent diagnostic method for simulation verification results based on knowledge graphs according to claim 1, characterized in that, In step S14, the knowledge graph is also equipped with an update mechanism. New simulation verification data, fault cases, and industry standard update data are collected regularly. The knowledge graph is updated with nodes and edges through incremental learning algorithms. During the update process, the new data is preprocessed first, and then conflict detection is performed with the existing knowledge graph. If there is a data conflict, the expert review result shall be used to determine whether to update, so as to ensure the timeliness and accuracy of the knowledge graph.

9. The intelligent diagnostic method for simulation verification results based on knowledge graphs according to claim 1, characterized in that, It also includes step S7: establishing a diagnostic result feedback mechanism, storing the final results of each diagnosis and the corresponding simulation verification data and feature data in the historical database, periodically analyzing the data in the historical database, mining new fault modes and diagnostic rules, adding the new fault modes as new nodes in the "fault mode" class and the new diagnostic rules as new nodes in the "diagnostic rule" class to the knowledge graph, so as to realize the self-improvement of the knowledge graph.

10. A knowledge graph-based intelligent diagnostic system for simulation verification results, characterized in that, The system for implementing the method of any one of claims 1-9 comprises: The knowledge graph construction module is used to collect data from the simulation verification domain, perform data preprocessing, ontology modeling, and graph database storage, and construct and update the knowledge graph of the simulation verification domain. The module includes a data acquisition unit, a data preprocessing unit, an ontology modeling unit, a graph storage unit, and a graph update unit. The data acquisition unit is used to collect data such as design documents and historical reports. The data preprocessing unit performs data cleaning and standardization. The ontology modeling unit defines concept classes and relationships. The graph storage unit uses a graph database to store graph data. The graph update unit realizes incremental updates of the graph. The simulation data acquisition module is used to acquire the simulation verification result data to be diagnosed. It includes a distributed data acquisition unit, a data encryption unit, and a data identification unit. The distributed data acquisition unit collects data from the simulation node, the data encryption unit encrypts the data using the AES-256 algorithm, and the data identification unit adds a timestamp and source identifier. The feature extraction module is used to extract features from the simulation verification result data. It includes a temporal feature extraction unit, a non-temporal feature extraction unit, and a feature fusion unit. The temporal feature extraction unit uses wavelet transform and LSTM network to extract temporal features. The non-temporal feature extraction unit uses fuzzy mathematics to extract attribute features. The feature fusion unit fuses different types of features to obtain feature data. The graph matching module is used to match feature data with knowledge graphs. It includes attribute matching unit, threshold matching unit and semantic matching unit, which respectively realize object localization, abnormal parameter marking and semantic association matching. The diagnostic reasoning module is used to determine the fault mode based on abnormal feature data. It includes a rule reasoning unit, a graph neural network reasoning unit, and a matching degree calculation unit. The rule reasoning unit filters fault modes based on diagnostic rules, the graph neural network reasoning unit calculates the fault mode score through the GCN network, and the matching degree calculation unit determines the preliminary diagnostic result. The result verification and output module is used to verify the preliminary diagnosis results and output the final results and solutions. It includes a result verification unit, a solution generation unit, and a result feedback unit. The result verification unit performs secondary comparison and verification, the solution generation unit outputs the fault solution, and the result feedback unit feeds back the diagnosis data to the historical database. The historical database module is used to store historical diagnostic data. It includes a data storage unit and a data analysis unit. The data storage unit stores diagnostic results, simulation data, and feature data. The data analysis unit mines new fault modes and diagnostic rules, providing data support for the improvement of the knowledge graph.

Images