CN-121980760-A - Knowledge graph-based intelligent diagnosis method and system for simulation verification result

CN121980760ACN 121980760 ACN121980760 ACN 121980760ACN-121980760-A

Abstract

The invention discloses a knowledge graph-based intelligent diagnosis method and system for simulation verification results, and relates to the technical field of intersection of simulation verification technology, intelligent diagnosis technology and knowledge graph technology. The method comprises the steps of firstly constructing a domain knowledge graph comprising simulation object attributes, verification standard thresholds, fault modes and diagnosis rules, then obtaining simulation verification data to be diagnosed, extracting time sequence/non-time sequence characteristics, associating characteristic data with the knowledge graph through a hierarchical matching strategy, screening abnormal characteristics, determining a primary fault mode by adopting a mixed mechanism of rule reasoning and graph neural network reasoning, outputting a final diagnosis result and a solution through secondary verification, and simultaneously establishing a knowledge graph updating and diagnosis result feedback mechanism. The problems that the traditional simulation diagnosis depends on manpower, is low in efficiency and is difficult to reuse in knowledge are solved, fault positioning accuracy and diagnosis efficiency are remarkably improved, a distributed simulation scene can be adapted, and good expansibility and timeliness are achieved.

Inventors

ZHANG YIJUN
ZHENG KOUQUAN
JING FENG
SHI YOUWEI
XING LIPENG
XIN DAN
ZHAO LE

Assignees

中国人民解放军国防科技大学

Dates

Publication Date: 20260505
Application Date: 20251224

Claims (10)

1. The intelligent diagnosis method of the simulation verification result based on the knowledge graph is characterized by comprising the following steps of: S1, constructing a knowledge graph in the simulation verification field, wherein the knowledge graph comprises simulation object attribute information, simulation verification standard threshold information, simulation fault mode information and fault diagnosis rule information, the simulation object attribute information comprises object structure parameters and performance parameters, the simulation verification standard threshold information is the qualification range of each performance parameter under different simulation scenes, the simulation fault mode information corresponds to the fault type when different performance parameters are abnormal, and the fault diagnosis rule information is the association logic of the performance parameter abnormality and the fault mode; S2, obtaining simulation verification result data to be diagnosed, wherein the simulation verification result data comprise time sequence data of each performance parameter, simulation scene parameters and simulation object identification information acquired in a simulation process; s3, extracting features of the simulation verification result data to obtain feature data, wherein if the simulation verification result data is time sequence data, the extracted time sequence features comprise data mean values, variances, peaks, valleys and trend change rates, and if the simulation verification result data is non-time sequence data, the extracted attribute features comprise deviation amounts of actual parameter values and preset reference values; s4, matching the feature data with the simulation verification field knowledge graph constructed in the step S1, positioning corresponding simulation object nodes in the knowledge graph based on the simulation object identification information, comparing the extracted feature data with simulation verification standard threshold information associated with the nodes, and screening out abnormal feature data exceeding a standard threshold range; S5, reasoning is carried out based on the abnormal feature data and fault diagnosis rule information in the knowledge graph, a corresponding simulation fault mode is determined, if a plurality of abnormal feature data exist in the reasoning process, the matching degree of each fault mode is calculated through the association weight of the fault mode and the plurality of abnormal features in the knowledge graph, and the fault mode with the highest matching degree is selected as a primary diagnosis result; S6, verifying the primary diagnosis result, calling typical feature data corresponding to the fault mode in the knowledge graph, performing secondary comparison with abnormal feature data of a simulation verification result to be diagnosed, outputting the fault mode as a final diagnosis result if the feature matching degree of the two feature data is greater than a preset threshold value of 85%, and returning to the step S3 to re-optimize the feature extraction dimension and then executing the subsequent steps again if the feature matching degree is less than or equal to 85%.
2. The intelligent diagnosis method of the simulation verification result based on the knowledge graph according to claim 1, wherein the specific process of constructing the knowledge graph in the simulation verification field in step S1 includes: S11, acquiring simulation verification field data, wherein the data sources comprise simulation object design documents, historical simulation verification reports, industry standard specifications and expert experience data, and the historical simulation verification reports comprise fault cases in past simulation and corresponding diagnosis results; S12, preprocessing the acquired data, including data cleaning to remove redundant error data, data standardization to unify parameter units and formats, and data structuring to convert unstructured text data into a key value pair format; s13, carrying out body modeling based on the preprocessed data, defining a core concept class of a knowledge graph, wherein the core concept class comprises a simulation object, performance parameters, verification standards, fault modes and diagnosis rules, and determining association relations among classes, wherein the association relations comprise an object, a parameter, a corresponding standard, a parameter abnormality, triggering, a fault and a fault-following rule; s14, storing the ontology model and the associated data by adopting a graph database, and constructing nodes and edges of the knowledge graph, wherein the node attributes comprise specific attribute values of the classes, the edge attributes comprise weights and confidence degrees of the associated relations, and an index mechanism of the knowledge graph is established for rapidly positioning the simulation object nodes and the associated information.
3. The intelligent diagnosis method of simulation verification result based on knowledge graph according to claim 1, wherein when the simulation verification result data to be diagnosed is obtained in the 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, data security protection is performed by adopting an encryption algorithm in the data transmission process, the encryption algorithm is an AES-256 algorithm, and a time stamp and a data source identifier are added to the obtained data for tracing the data acquisition node and the acquisition time.
4. The intelligent diagnosis method of the simulation verification result based on the knowledge graph according to claim 1, wherein the feature extraction in the step S3 adopts a fusion algorithm, for time sequence performance parameter data, wavelet transformation is firstly adopted to remove data noise, then LSTM network is adopted to extract time sequence deep features, the deep features are fused with traditional statistical features to form final time sequence feature data, and for non-time sequence data, fuzzy mathematical method is adopted to fuzzify parameter deviation to obtain fuzzy feature vectors as attribute feature data.
5. The intelligent diagnosis method of the simulation verification result based on the knowledge graph according to claim 1, wherein in the step S4, a hierarchical matching strategy is adopted for matching the feature data with the knowledge graph, the first layer is attribute matching, the identification information of the simulation object is precisely matched with the attribute of the 'simulation object' class node in the knowledge graph, the target node is positioned, the second layer is threshold matching, interval matching is carried out on 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, abnormal parameters exceeding the interval are marked, the third layer is semantic matching, semantic association is carried out on the abnormal parameters and the 'performance parameter' class node through the semantic relation in the knowledge graph, and the matching accuracy is ensured.
6. The intelligent diagnosis method of the simulation verification result based on the knowledge graph according to claim 1, wherein the reasoning process in the step S5 adopts a hybrid reasoning mechanism, rule-based reasoning and graph neural network-based reasoning are combined, rule-based reasoning utilizes association logic stored by 'diagnosis rule' class nodes in the knowledge graph to primarily screen fault modes, the graph neural network-based reasoning uses the knowledge graph as a topological structure, abnormal characteristic data is used as node input characteristics, the association weights among nodes are learned through a GCN network, the score of each fault mode node is calculated, the fault mode with the highest score is the primary diagnosis result, and the association weights are obtained through historical fault case data training.
7. The intelligent diagnosis method of simulation verification result based on knowledge graph according to claim 1, wherein the verification process in step S6 further comprises calling the "solution" information corresponding to the failure mode in the knowledge graph, if the final diagnosis result is determined, synchronously outputting the corresponding failure solution, the solution includes a failure checking step, parameter adjustment suggestion and simulation system optimization direction, and if verification fails to execute step S3 again, increasing feature extraction index of the dimension based on unmatched feature dimension, such as increasing periodic feature of time series data or deviation change rate feature of non-time series data.
8. The intelligent diagnosis method of the simulation verification result based on the knowledge graph according to claim 1, wherein the knowledge graph in the step S14 is further provided with an updating mechanism, new simulation verification data, fault cases and industry standard updating data are collected periodically, nodes and edges of the knowledge graph are updated through an incremental learning algorithm, the new data are preprocessed in the updating process, then conflict detection is carried out on the new data and the existing knowledge graph, if data conflict exists, whether the updating is carried out is determined based on expert auditing results, and timeliness and accuracy of the knowledge graph are ensured.
9. The intelligent diagnosis method of simulation verification results based on a knowledge graph according to claim 1, further comprising the step S7 of establishing a diagnosis result feedback mechanism, storing the final result of each diagnosis, the corresponding simulation verification data and the feature data into a historical database, analyzing the data in the historical database regularly, mining new fault modes and diagnosis rules, adding the new fault modes as new nodes of the 'fault modes' and the new diagnosis rules as new nodes of the 'diagnosis rules' to the knowledge graph, and realizing self-improvement of the knowledge graph.
10. A knowledge-graph-based intelligent diagnostic system for simulating verification results, characterized in that it is configured to implement the method of any one of claims 1-9, the system comprising: The knowledge graph construction module is used for collecting data in the simulation verification field, carrying out data preprocessing, body modeling and graph database storage, constructing and updating a knowledge graph in the simulation verification field, and comprises a data collection unit, a data preprocessing unit, a body modeling unit, a graph storage unit and a graph updating unit, wherein the data collection unit is used for collecting data such as design documents, historical reports and the like, the data preprocessing unit carries out data cleaning standardization, the body modeling unit defines concept types and association relations, the graph storage unit stores graph data by adopting the graph database, and the graph updating unit realizes graph increment updating; The simulation data acquisition module is used for acquiring simulation verification result data to be diagnosed and comprises a distributed data acquisition unit, a data encryption unit and a data identification unit, wherein the distributed data acquisition unit acquires data from a simulation node, the data encryption unit encrypts the data by adopting an AES-256 algorithm, and the data identification unit adds a time stamp and a source identification; The feature extraction module is used for extracting features of simulation verification result data and comprises a time sequence feature extraction unit, a non-time sequence feature extraction unit and a feature fusion unit, wherein the time sequence feature extraction unit adopts wavelet transformation and an LSTM network to extract time sequence features, the non-time sequence feature extraction unit adopts a fuzzy mathematical method to extract attribute features, and the feature fusion unit fuses different types of features to obtain feature data; the map matching module is used for matching the feature data with the knowledge map and comprises an attribute matching unit, a threshold matching unit and a semantic matching unit, and object positioning, abnormal parameter marking and semantic association matching are respectively realized; The diagnosis reasoning module is used for reasoning and determining a fault mode based on the abnormal characteristic data and comprises a rule reasoning unit, a graph neural network reasoning unit and a matching degree calculating unit, wherein the rule reasoning unit screens the fault mode based on diagnosis rules, the graph neural network reasoning unit calculates a fault mode score through a GCN (global carrier wave) network, and the matching degree calculating unit determines a preliminary diagnosis result; the result verification and output module is used for verifying the primary diagnosis result and outputting a final result and a solution, and comprises a result verification unit, a scheme generation unit and a result feedback unit, wherein the result verification unit performs secondary comparison verification, the scheme generation unit outputs a fault solution, and the result feedback unit feeds back diagnosis data to the historical database; The historical database module is used for storing diagnosis historical data and comprises a data storage unit and a data analysis unit, wherein the data storage unit stores diagnosis results, simulation data and characteristic data, and the data analysis unit is used for mining new fault modes and diagnosis rules and providing data support for knowledge graph improvement.

Description

Knowledge graph-based intelligent diagnosis method and system for simulation verification result Technical Field The invention relates to the technical field of intersection of a simulation verification technology, an intelligent diagnosis technology and a knowledge graph technology, in particular to a simulation verification result intelligent diagnosis method and system based on a knowledge graph. Background In the research and development process of a complex system, simulation verification is a key link for verifying the design rationality of the system and troubleshooting potential faults. With the increase of the complexity of the system, the data volume generated by simulation verification increases exponentially, and the simulation verification relates to multiple types of performance parameters and multiple simulation scenes. The traditional simulation verification result diagnosis method has the obvious defects that an existing diagnosis multi-dependency engineer manually compares simulation data with a standard threshold value, analyzes association of abnormal parameters and faults, consumes a large amount of time for checking for multi-parameter concurrent abnormal scenes, is difficult to meet real-time diagnosis requirements, the diagnosis experience of the engineer is stored in a document form, a structural organization is lacked, a new engineer needs long-term learning to master, fault diagnosis knowledge among different projects is difficult to share, the traditional method lacks deep mining of parameter association relations, fault sources are easy to misjudge when a plurality of parameters are abnormal, the existing data acquisition and diagnosis method is difficult to realize cross-node data integration and lacks a data safety protection mechanism for a distributed simulation system, and the existing diagnosis system is difficult to quickly incorporate the new knowledge when an industry standard iteration and a new fault mode appear, so that the diagnosis result and the actual requirement are disjointed. The existing diagnosis is based on manual comparison of simulation data and standard threshold values by multiple engineers, a large amount of time is required to be consumed for investigation, real-time diagnosis needs are difficult to meet, the diagnosis experience of the engineers is stored in a document form, a new engineer can learn for a long time and is difficult to share fault diagnosis knowledge, the traditional method lacks deep excavation of parameter association relations, when a plurality of parameters are abnormal, the fault sources are easy to misjudge, for a distributed simulation system, a data safety protection mechanism is lacking, and when industry standard iteration and a new fault mode occur, the existing diagnosis system is difficult to quickly incorporate new knowledge, so that a diagnosis result is disjointed with the actual needs. Disclosure of Invention The invention aims to provide an intelligent diagnosis method and an intelligent diagnosis system for a simulation verification result based on a knowledge graph, which are used for solving the technical problems of low efficiency, difficult multiplexing of knowledge, inaccurate positioning, poor suitability and delayed updating in the conventional simulation verification result diagnosis in the background art, realizing automation, intellectualization and expandability of the simulation diagnosis and simultaneously guaranteeing the data safety and the real-time updating of knowledge in a distributed scene. In order to achieve the purpose, the invention provides the following technical scheme that the intelligent diagnosis method and the system for the simulation verification result based on the knowledge graph comprise the following steps: S1, constructing a knowledge graph in the simulation verification field, wherein the knowledge graph comprises simulation object attribute information, simulation verification standard threshold information, simulation fault mode information and fault diagnosis rule information, the simulation object attribute information comprises object structure parameters and performance parameters, the simulation verification standard threshold information is the qualification range of each performance parameter under different simulation scenes, the simulation fault mode information corresponds to the fault type when different performance parameters are abnormal, and the fault diagnosis rule information is the association logic of the performance parameter abnormality and the fault mode; S2, obtaining simulation verification result data to be diagnosed, wherein the simulation verification result data comprise time sequence data of each performance parameter, simulation scene parameters and simulation object identification information acquired in a simulation process; s3, extracting features of the simulation verification result data to obtain feature data, wherein if the simulation verif