A model validation method based on SysML language

By constructing a unified semantic attribute graph across views and integrating specific algorithms, the problems of view fragmentation and semantic inconsistency in SysML modeling are solved, achieving efficient model consistency and intelligent verification, and improving modeling quality and efficiency.

CN123114436APending Publication Date: 2026-11-17HANGZHOU HUAWANG SYST TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611304180.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-26
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing SysML modeling methods lack cross-view semantic relationship modeling means, making it difficult to link information between multiple views and ensuring model consistency. Traditional tools cannot adapt to the dynamic expansion of verification rules in different system architectures or industry fields, and lack intelligent feedback capabilities, resulting in low efficiency and high repair costs for designers.

Method used

Construct a unified semantic attribute graph across views, integrating the VF2++ subgraph isomorphism algorithm, the GTrie graph pattern enumeration algorithm, and the Rete-NT incremental rule engine to achieve structural consistency verification, semantic integrity verification, and defect root cause localization among multiple views, supporting rule scalability and engineering integration.

Benefits of technology

It enables information linkage between multi-view modeling, improves the accuracy and traceability of model consistency analysis, has efficient structural conflict identification capabilities and dynamic rule loading capabilities, reduces the analysis and repair costs for designers, and improves modeling quality and efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123114436A_ABST
    Figure CN123114436A_ABST
Patent Text Reader

Abstract

This invention discloses a model validation method based on the SysML language, comprising the following steps: S1, obtaining and parsing the SysML model file to generate a set of model elements; S2, inputting the set of model elements into the semantic modeling module to construct a semantic attribute graph; S3, using the semantic attribute graph as input, applying the VF2++ subgraph isomorphism algorithm to output structural consistency validation results; S4, based on the structural consistency validation results, calling the enumeration algorithm to output structural integrity validation results; S5, using the structural integrity validation results and the semantic attribute graph as joint input to output normativity validation results; S6, summarizing all validation results to generate a unified validation report; S7, using the validation report as input, executing the minimum cut set generation algorithm to output a complete defect location and repair solution. This invention integrates the VF2++ subgraph isomorphism algorithm and Rete-NT incremental rules to achieve model validation based on the SysML language.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of system modeling and systems engineering language technology, and in particular to a model validation method based on the SysML language. Background Technology

[0002] With the rapid development of complex systems, embedded systems, and large-scale collaborative systems, model-driven engineering is increasingly widely used in the field of systems engineering. SysML, as a standardized modeling tool, is widely used to describe multi-dimensional views of system requirements, structure, behavior, and constraints. SysML supports modeling systems using various views such as requirement diagrams, block definition diagrams, internal block diagrams, state diagrams, and activity diagrams. Its multi-view coexistence feature allows system modeling to more comprehensively and flexibly reflect the actual design intent. However, in modeling practice, due to the lack of standardized cross-view semantic relationship modeling methods, multiple views often appear isolated, information is difficult to link, and model consistency is difficult to guarantee.

[0003] Existing model validation methods mostly focus on static checks of formal rules such as structure, syntax, and naming conventions within a single view, lacking in-depth analysis of the overall structural consistency and design intent integrity of the model. For example, performing static rule matching checks only on activity diagrams or state diagrams may fail to identify conflicts in control logic and data flow dependencies between these behavioral views and structural views. Furthermore, traditional model validation tools mostly employ hard-coded rule bases or static XML validation methods, lacking general reasoning capabilities and structural context awareness, and cannot adapt to the dynamic expansion requirements of validation rules in different system architectures or industry sectors. Regarding validation result output, traditional tools typically only return the IDs and text descriptions of model elements that do not meet the rules, lacking analysis of problem causes, defect structure localization, and remediation suggestions. This forces designers to manually analyze the error context, resulting in low efficiency, high modification costs, and a high risk of introducing new design flaws, creating an inefficient cycle of "fix—re-error—re-fix."

[0004] To address the aforementioned issues, this invention proposes a model validation method based on the SysML language. It constructs a unified semantic attribute graph across views and integrates the VF2++ subgraph isomorphism algorithm, the GTrie graph pattern enumeration algorithm, the Rete-NT incremental rule engine, and the minimum cut set analysis method to form an intelligent validation mechanism for the entire system modeling process. Through structure-level graph matching, template structure recognition, rule reasoning, and path minimization analysis, this method can achieve structural consistency verification, semantic integrity verification, and defect root cause localization across multiple views, significantly improving system modeling quality and validation efficiency. This method is not only applicable to standard SysML modeling environments but also possesses excellent rule scalability and engineering integration capabilities, enabling it to meet diverse needs for modeling specifications, semantic relationships, and validation feedback in different industry contexts.

[0005] Therefore, compared with the existing technology which can only perform single-view rule checks and lacks intelligent feedback capabilities and defect location mechanisms, the method provided by this invention can effectively solve a series of core problems such as fragmented modeling views, semantic inconsistency, coarse verification granularity, and low repair efficiency, providing a more practical and intelligent solution for model verification of complex systems.

[0006] Therefore, how to provide a model validation method based on the SysML language is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] One objective of this invention is to propose a model validation method based on the SysML language. This invention provides an intelligent model validation method based on SysML, which constructs a unified semantic attribute graph and integrates the VF2++ subgraph isomorphism algorithm, structural enumeration technology, and the Rete-NT incremental rule engine to achieve multi-dimensional automatic validation of the system model in terms of structural consistency, design integrity, and standardization. This method supports cross-view model consistency analysis, has defect location and repair suggestion generation functions, and improves the automation, intelligence, and engineering practicality of the SysML modeling process.

[0008] A model validation method based on SysML language according to an embodiment of the present invention includes the following steps:

[0009] S1. Obtain the SysML model file to be verified and parse it to extract model elements and metadata, and generate a set of model elements containing multi-view element information.

[0010] S2. Input the set of model elements into the semantic modeling module and construct a unified semantic attribute graph according to the preset semantic mapping rules;

[0011] S3. Using the semantic attribute graph as input, apply the VF2++ subgraph isomorphism algorithm to perform matching analysis on the structural subgraphs in the semantic attribute graph, identify structural consistency conflicts or missing relationships between cross-views, and output the structural consistency verification results.

[0012] S4. Based on the structural consistency verification result, the GTrie graph pattern enumeration algorithm is further called to perform structural pattern recognition on the node combination to determine whether there is a missing, deformed or redundant design structural template, and output the structural integrity verification result.

[0013] S5. Take the structural integrity verification result and the semantic attribute graph as joint input, load the rule base and start the Rete-NT incremental rule engine, perform verification according to the structural rules, behavioral rules and user-defined rules in the rule base, and output the normative verification result.

[0014] S6. Summarize the results of structural consistency verification, structural integrity verification, and conformity verification to generate a unified structural verification report;

[0015] S7. Taking the structural verification report as input, execute the minimum cut set generation algorithm to determine the minimum set of model elements that cause the defect based on the dependency path between defective elements, and generate corresponding repair suggestions according to the rule base, and output a complete defect location and repair plan.

[0016] Optionally, S1 specifically includes:

[0017] S11. Load the SysML model file to be verified, identify the storage format of the SysML model file, including XMI and XML formats, and read the model file hierarchically through a parser based on document object model and event-driven model, identify the package structure, model segmentation and the starting position of various views in the model, and provide an index structure for model element extraction.

[0018] S12. Based on the index structure, extract model elements and three types of model view content from the model file as needed, including extracting requirement elements and derived links from the requirement view, extracting Block, Port and Connector component elements from the structure view, and extracting Activity, Action, ControlFlow, State and Transition behavior-related elements from the behavior view. At the same time, record the level of the model view content, parent-child relationship and cross-graph reference node path.

[0019] S13. Call the metadata extraction module to read the metadata information of each model element. The metadata information includes the element's unique identifier, name, type, view, attribute fields, binding constraints, inheritance relationship, implementation path, call dependency, and external references. The attribute fields are read in key-value pair structure, the constraints are extracted in the form of syntax expressions to extract symbols, variables, and target scope, and the dependency and reference information are obtained through the internal reference path defined by the modeling tool.

[0020] S14. The extracted model elements and corresponding metadata are structurally encapsulated and divided into requirement class, structure class and behavior class model element sets. The elements in the model element set are subjected to integrity verification and duplicate identification filtering to remove missing attributes and illegal nodes, and finally a model element set with a unified format is generated.

[0021] Optionally, S2 specifically includes:

[0022] S21. Based on the received model element set, which consists of multiple model elements, each model element contains a unique identifier, type, attribute information, constraints, and relationship information with other model elements. The semantic mapping rule library is called to parse the attribute fields and relationship fields of the model element set. According to the semantic mapping rules, the different types of requirement elements, structural elements, and behavioral elements are mapped to a unified data representation model.

[0023] S22. Represent each model element as a node in a graph structure, and establish semantic connection edges between these nodes based on the various relationships between the model elements, including calling relationships, dependency relationships, composition relationships, referencing relationships, and transformation relationships, to form an attribute graph containing a set of nodes and a set of edges.

[0024] S23. Perform structural optimization processing on the completed attribute graph, supplement the view source type, model semantic label and cross-view matching identifier of each node, add semantic weight information to all edges, and form a semantic attribute graph.

[0025] Optionally, S3 specifically includes:

[0026] S31. Using the constructed semantic attribute graph as the input graph, read the structural information of all nodes and edges in the graph. Nodes are used to represent various elements in the model, including requirement blocks, structures, state nodes, and behavior modules. Edges represent the semantic relationships of calls, dependencies, state transitions, and connections between them. At the same time, extract the view source, type label, and naming attributes of each node, as well as the semantic identifier and direction information of each edge.

[0027] S32. Load a target structure subgraph from the structure pattern library as a reference template for consistency verification. Each node in the target structure subgraph represents a model component, and each edge represents the structural relationship that the components must satisfy, with type labels and semantic connection requirements, forming a standardized structure template.

[0028] S33. In the semantic attribute graph, call the VF2++ subgraph isomorphism algorithm to perform structure matching on the standardized structure template, and construct a mapping function from the template subgraph to the input graph. This ensures that for any pair of connected nodes in the template graph And it satisfies the mapping verification condition, the expression of which is:

[0029] ;

[0030] in, , This represents two connected nodes in the template diagram. This represents the mapping result of nodes in the semantic attribute graph. The set of edges representing the semantic attribute graph. Indicates the template node type label, Indicates a node type tagging function. Semantic labels representing edges in a structural template graph;

[0031] S34. Identify nodes and edges that pass the mapping verification conditions, output the structural consistency verification results, and locate nodes and edges that fail the mapping verification conditions as areas of structural inconsistency. Record the view type, node name, edge connection type and upstream and downstream relationship as structural conflict information to provide to the subsequent defect analysis module for generating model verification reports and structural repair suggestions.

[0032] 5. The model validation method based on SysML language according to claim 1, wherein step S4 specifically includes:

[0033] S41. Based on the structural consistency verification results, extract all regions with structural conflicts in the semantic attribute graph, locate the subgraph range corresponding to each conflict region, and extract each range as a local subgraph containing a set of nodes and a set of edges. Combine all local subgraphs into a candidate subgraph set. Each candidate subgraph contains model elements in the view and their semantic connection relationships, which are used for the integrity enumeration analysis of graph structure patterns.

[0034] S42. Load multiple structural template graphs from the structural pattern library. Each structural template graph contains a subgraph structure composed of a set of model elements with specific semantic connections. It contains clear node type labels and edge label definitions. Based on the candidate subgraph set, construct all structural template graphs into a GTrie index tree by sharing graph prefixes. Establish the index mapping relationship between node topology paths and structural patterns to form a set of structural template indexes for enumeration and traversal. Cache the index tree for fast matching calculation.

[0035] S43. For each subgraph in the candidate subgraph set, call the target structure template graph from the GTrie structure in turn, perform the structure template enumeration operation in the current subgraph by graph traversal, and define the structure pattern matching function.

[0036] S44. Based on the graph structure enumeration matching results, record the templates that are successfully matched, the templates that are not successfully matched, the template instances whose structures in the candidate subgraphs exceed the template definition range, and the template instances whose semantic connections or node types are inconsistent. Classify these four types of information into normal items, missing items, redundant items, and deformed items, and integrate them into a set of structural integrity verification results. This set includes the structural template identifier, associated model element number, path information of abnormal structures, and corresponding view range for each type of result, and serves as the structural integrity verification result.

[0037] 6. The model validation method based on SysML language according to claim 1, wherein step S5 specifically includes:

[0038] S51. Import the structural consistency verification result, structural integrity verification result, and semantic attribute graph into the verification module as input, load the rule set, and define each rule as a rule structure composed of condition trigger body and response action;

[0039] S52. Initialize the matching of all model nodes in the semantic attribute graph using a rule-based matching initialization function. Construct a rule-based pattern matching index based on node attributes and edge relationships. Extract node combinations through matching patterns and generate conflict sets by combining defect information with the index structure. The conflict set generation function is defined as follows:

[0040] ;

[0041] in This represents the process by which the matching engine solves for the joint conditions between the rules and the model. This is a function for judging the consistency of semantic structure. For a set of rules, For semantic attribute graphs, For the k-th rule in the rule set, A collection of structural defect information;

[0042] S53. Traverse all rules that meet the conditions in the conflict set, add the matching model structure combination to the trigger queue, and perform structural relationship judgment, label verification and model state detection according to the rule action content. Extract the execution results and convert them into semantic feedback actions.

[0043] S54. Summarize the execution records of all triggered rules, and generate normative verification results based on semantic feedback actions. The results include the rule hit identifier, the corresponding model element number, the logical hit path, and the verification category marker.

[0044] 7. The model validation method based on SysML language according to claim 1, wherein step S6 specifically includes:

[0045] S61. Receive the structural consistency verification result, structural integrity verification result, and standardization verification result. Number the defect items contained in each type of verification result in a unified manner. Based on the model nodes and connection relationships in the semantic attribute graph, identify the model elements involved in each defect path, establish a mapping index table between defect number and model element number, and mark the dependency relationship between defect path and model structure.

[0046] S62. Based on the established mapping index table, analyze the node overlap between defect paths, divide defect items with structural overlap or connection coupling into the same structural cluster group, and each cluster group forms a structural verification result fragment, which includes the aggregated model element set, semantic path range, the model view to which it belongs and the trigger rule record, and generate a cluster group list.

[0047] S63. Perform structural mapping analysis on all cluster group lists, extract the distribution relationship of each group of model elements in the semantic attribute graph, identify cross-connection paths and merge duplicate structures, construct a structural influence map under multiple views, complete the dual alignment of defect clustering and view mapping, form a structural influence view and generate a unified representation format.

[0048] S64. Based on the path information, rule sources, structural influence areas and corresponding model element indexes in the structural influence view and model diagram, a unified structural verification report is generated. The report includes defect type classification, distribution statistics, cross-view distribution analysis, structural influence summary and rule trigger statistics.

[0049] 8. The model validation method based on SysML language according to claim 1, wherein step S7 specifically includes:

[0050] S71. Using the structural verification report as input, construct a directed graph where nodes represent model elements and edges represent structural dependencies and defect propagation paths between elements. Label the initial fault source nodes based on the defect result set and form a defect propagation graph.

[0051] S72. Perform a minimum cut set search on the defect propagation graph to extract the minimum combination of nodes that covers all defect paths. The minimum cut set search expression is:

[0052] ;

[0053] in, For the set of minimal cut sets, , Both represent a single minimal cut set. To verify the set of nodes that affect graph a, For the first Defect path, This is a set of structural integrity verification results. This is a set of normative verification results. Representing cut set With defect path The intersection of the nodes is not an empty set. This indicates that there is no smaller cut set that can cover the same defect path;

[0054] S73. Calculate the total repair cost of each minimal cut set containing nodes, sort all minimal cut sets according to the cost function results, select the combination with the lowest cost as the priority repair candidate, and extract the model elements and their respective view and semantic path information contained in each group.

[0055] S74. Based on the sorted minimum cut set sequence, match the repair strategy template corresponding to the node type in the rule base, construct targeted repair suggestions, and output the defect location index, minimum cut set detailed mapping table, and structural repair suggestions.

[0056] The beneficial effects of this invention are:

[0057] 1. The model validation method based on SysML language proposed in this invention has significant beneficial effects on solving the consistency and integrity analysis of multi-view models. First, by constructing a semantic attribute graph, this method realizes a unified expression and semantic association between multiple modeling views such as requirements, structure, and behavior, effectively breaking down information barriers between views, making up for the problems of model fragmentation and semantic disconnect in existing technologies, and significantly improving the accuracy and traceability of cross-view modeling consistency analysis.

[0058] 2. Secondly, the method incorporates the VF2++ subgraph isomorphism algorithm and the GTrie graph pattern enumeration algorithm, enabling structural consistency and templated structure recognition to achieve high efficiency and coverage at the graph computation level. This allows for the accurate detection of structural conflicts, deformations, and omissions in the model, overcoming the technical bottleneck of traditional rule-based verification's inability to detect hidden structural problems. Simultaneously, the integration of the Rete-NT incremental rule engine enables dynamic rule loading and high-frequency inference capabilities, allowing users to quickly build customized verification logic under different project contexts, enhancing the scalability and flexible adaptability of the verification rules.

[0059] 3. Furthermore, this invention proposes a minimum cut set analysis algorithm based on defect propagation paths. This algorithm can identify the minimum impact range that causes errors in complex model dependencies and generate targeted repair suggestions by combining them with a rule base. This constructs a closed-loop feedback mechanism of "verification-location-suggestion," significantly reducing the cost of problem analysis and correction for designers and improving the intelligence level of model quality assurance. In summary, this invention has achieved breakthroughs in modeling semantic fusion, intelligent structural recognition, rule extensibility, and defect location capabilities, providing a more practical model verification solution for the modeling process of complex systems. Attached Figure Description

[0060] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0061] Figure 1 This is a flowchart of a model validation method based on the SysML language proposed in this invention;

[0062] Figure 2 This is a schematic diagram of the algorithm structure of a model validation method based on SysML language proposed in this invention. Detailed Implementation

[0063] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0064] refer to Figure 1-2 A model validation method based on the SysML language includes the following steps:

[0065] S1. Obtain the SysML model file to be verified and parse it to extract model elements and metadata, and generate a set of model elements containing multi-view element information.

[0066] S2. Input the set of model elements into the semantic modeling module and construct a unified semantic attribute graph according to the preset semantic mapping rules;

[0067] S3. Using the semantic attribute graph as input, apply the VF2++ subgraph isomorphism algorithm to perform matching analysis on the structural subgraphs in the semantic attribute graph, identify structural consistency conflicts or missing relationships between cross-views, and output the structural consistency verification results.

[0068] S4. Based on the structural consistency verification result, the GTrie graph pattern enumeration algorithm is further called to perform structural pattern recognition on the node combination to determine whether there is a missing, deformed or redundant design structural template, and output the structural integrity verification result.

[0069] S5. Take the structural integrity verification result and the semantic attribute graph as joint input, load the rule base and start the Rete-NT incremental rule engine, perform verification according to the structural rules, behavioral rules and user-defined rules in the rule base, and output the normative verification result.

[0070] S6. Summarize the results of structural consistency verification, structural integrity verification, and conformity verification to generate a unified structural verification report;

[0071] S7. Taking the structural verification report as input, execute the minimum cut set generation algorithm to determine the minimum set of model elements that cause the defect based on the dependency path between defective elements, and generate corresponding repair suggestions according to the rule base, and output a complete defect location and repair plan.

[0072] This invention proposes a SysML model verification method that integrates semantic modeling and graph computation technologies. Addressing the problems of fragmented modeling views, coarse verification granularity, and lack of intelligent feedback in existing methods, it achieves deep integration of semantic relationships between multi-view modeling elements by constructing a unified semantic attribute graph. The method integrates VF2++ subgraph isomorphism and GTrie graph pattern enumeration algorithms, enabling efficient identification of cross-view structural conflicts and design template anomalies, thus improving the depth and breadth of structural consistency and integrity verification. By introducing the Rete-NT incremental rule engine, the method supports dynamic structural and behavioral rule reasoning, achieving efficient execution of model normative verification. Furthermore, by combining minimum cut set analysis technology, it performs path-based tracking and minimum localization of model defects, automatically generating repair suggestions and constructing a closed-loop "verification-localization-repair" mechanism. Overall, this method possesses high scalability, high accuracy, and strong feedback capabilities, significantly improving the automation and intelligence level of the complex system modeling process.

[0073] In this embodiment, S1 specifically includes:

[0074] S11. Load the SysML model file to be verified, identify the storage format of the SysML model file, and support formats including XMI format and XML format; read the model file hierarchically through a parser based on document object model and event-driven model, identify the package structure, model segmentation and the starting position of various views in the model, and provide an index structure for model element extraction.

[0075] S12. Based on the index structure, extract model elements and three types of model view content from the model file as needed, including extracting requirement elements and derived links from the requirement view, extracting Block, Port and Connector component elements from the structure view, and extracting Activity, Action, ControlFlow, State and Transition behavior-related elements from the behavior view. At the same time, record the level of the model view content, parent-child relationship and cross-graph reference node path.

[0076] S13. Call the metadata extraction module to read the metadata information of each model element. The metadata information includes the element's unique identifier, name, type, view, attribute fields, binding constraints, inheritance relationship, implementation path, call dependency, and external references. The attribute fields are read in key-value pair structure, the constraints are extracted in the form of syntax expressions to extract symbols, variables, and target scope, and the dependency and reference information are obtained through the internal reference path defined by the modeling tool.

[0077] S14. The extracted model elements and corresponding metadata are structurally encapsulated and divided into requirement class, structure class and behavior class model element sets. The elements in the model element set are subjected to integrity verification and duplicate identification filtering to remove missing attributes and illegal nodes, and finally a model element set with a unified format is generated.

[0078] This step first uses a parsing mechanism compatible with formats such as XMI and XML to read the SysML model file hierarchically and construct a model content index structure. Then, it extracts model elements and reference paths from the three core views of requirements, structure, and behavior, completing the initial decomposition of the multi-view modeling content. The metadata extraction module extracts semantic information such as element identifiers, attributes, dependencies, and constraints, achieving in-depth semantic mining of the modeling. Finally, model elements from different sources and categories, along with their metadata, are structurally encapsulated to form a semantically consistent and formatted set of model elements, providing a complete and accurate input foundation for subsequent semantic modeling and graph computation analysis. This step ensures the comprehensiveness, accuracy, and semantic clarity of model information extraction and is the starting core of the entire model validation process.

[0079] In this embodiment, S2 specifically includes:

[0080] S21. Based on the received model element set, which consists of multiple model elements, each model element contains a unique identifier, its type, attribute information, constraints, and relationship information with other model elements; call the semantic mapping rule library to parse the attribute fields and relationship fields of the model element set, and map the different types of requirement elements, structural elements, and behavioral elements to a unified data representation model according to the semantic mapping rules;

[0081] S22. Represent each model element as a node in a graph structure, and establish semantic connection edges between these nodes based on the various relationships between the model elements, including calling relationships, dependency relationships, composition relationships, referencing relationships, and transformation relationships, to form an attribute graph containing a set of nodes and a set of edges.

[0082] S23. Perform structural optimization processing on the completed attribute graph, supplement the view source type, model semantic label and cross-view matching identifier of each node, add semantic weight information to all edges, and form a semantic attribute graph.

[0083] This step, based on the extracted set of model elements, uses pre-defined semantic mapping rules to uniformly transform the multi-view modeling elements, unifying various elements such as requirements, structures, and behaviors into a unified data representation model, and expressing the modeling semantics in a graph structure. By abstracting each model element as a node in the graph and establishing connections based on their calling, dependency, and combination semantic relationships, an initial attribute graph with semantic connectivity is constructed. Further structural optimization and semantic enhancement are performed on this graph, improving node source labeling and cross-view matching identifiers, and supplementing edge semantic weights. Finally, a semantic attribute graph with semantic consistency and cross-view expressive capabilities is generated, providing a rigorous data foundation for subsequent graph algorithm validation.

[0084] In this embodiment, S3 specifically includes:

[0085] S31. Using the constructed semantic attribute graph as the input graph, read the structural information of all nodes and edges in the graph. Nodes are used to represent various elements in the model, including requirement blocks, structures, state nodes and behavior modules. Edges represent the semantic relationships of calls, dependencies, state transitions and connections between them. At the same time, extract the view source, type label and naming attributes of each node, as well as the semantic identifier and direction information of each edge.

[0086] S32. Load a target structure subgraph from the structure pattern library as a reference template for consistency verification. Each node in the target structure subgraph represents a model component, and each edge represents the structural relationship that the components must satisfy, with type labels and semantic connection requirements, forming a standardized structure template.

[0087] S33. In the semantic attribute graph, call the VF2++ subgraph isomorphism algorithm to perform structure matching on the standardized structure template, and construct a mapping function from the template subgraph to the input graph. This ensures that for any pair of connected nodes in the template graph And it satisfies the mapping verification condition, the expression of which is:

[0088] ;

[0089] in, , This represents two connected nodes in the template diagram. This represents the mapping result of nodes in the semantic attribute graph. The set of edges representing the semantic attribute graph. Indicates the template node type label, Indicates a node type tagging function. Semantic labels representing edges in a structural template graph;

[0090] S34. Identify nodes and edges that pass the mapping verification conditions, output the structural consistency verification results, and locate nodes and edges that fail the mapping verification conditions as areas of structural inconsistency. Record the view type, node name, edge connection type and upstream and downstream relationship as structural conflict information to provide to the subsequent defect analysis module for generating model verification reports and structural repair suggestions.

[0091] This step, based on semantic attribute graphs and combined with standardized template structures from the structural pattern library, employs the VF2++ subgraph isomorphism algorithm to perform high-precision matching analysis of structural relationships in the model, achieving automatic verification of cross-view structural consistency. By mapping requirement, structural, and behavioral nodes in the SysML model to semantic nodes in the graph, and constructing a matching mapping function based on semantic connections, the system can identify the complete matching status and potential conflict points of structural templates in actual modeling. Nodes and edges that do not meet the mapping verification conditions are accurately marked as structural inconsistency areas, and detailed records are made based on their respective views, semantic connections, and location context, providing basic data support for subsequent defect localization and repair. This process effectively overcomes the problems of traditional models having single rules and coarse verification granularity, achieving graph-level modeling and algorithm-level identification of structural consistency problems, significantly improving the automation, interpretability, and practicality of model consistency verification.

[0092] In this embodiment, S4 specifically includes:

[0093] S41. Based on the structural consistency verification results, extract all regions with structural conflicts in the semantic attribute graph, locate the subgraph range corresponding to each conflict region, and extract each range as a local subgraph containing a set of nodes and a set of edges. Combine all local subgraphs into a candidate subgraph set. Each candidate subgraph contains model elements in the view and their semantic connection relationships, which are used for the integrity enumeration analysis of graph structure patterns.

[0094] S42. Load multiple structural template graphs from the structural pattern library. Each structural template graph contains a subgraph structure composed of a set of model elements with specific semantic connection relationships. It contains clear node type labels and edge label definitions. Based on the candidate subgraph set, construct all structural template graphs into a GTrie index tree by sharing graph prefixes. Establish the index mapping relationship between node topology paths and structural patterns to form a set of structural template indexes for enumeration and traversal. Cache the index tree for fast matching calculation.

[0095] S43. For each subgraph in the candidate subgraph set, sequentially call the target structure template graph from the GTrie structure, perform structure template enumeration operation in the current subgraph by graph traversal, and define a structure pattern matching function. The expression of the structure pattern matching function is as follows:

[0096] ;

[0097] in, This is a structural template diagram. As candidate subgraphs, For structural embedding mapping, This represents the set of all matching instances that satisfy the consistency of topology, node type, and connection semantics. This means mapping each node in the template graph to a node in the candidate subgraph. This indicates that the corresponding node in the candidate subgraph has an edge. The semantic labels represent the edges in the template graph. The semantic label represents the mapped edge in the candidate subgraph;

[0098] S44. Based on the graph structure enumeration matching results, record the templates that are successfully matched, the templates that are not successfully matched, the template instances whose structures in the candidate subgraphs exceed the template definition range, and the template instances whose semantic connections or node types are inconsistent. Classify these four types of information into normal items, missing items, redundant items, and deformed items, and integrate them into a set of structural integrity verification results. This set includes the structural template identifier, associated model element number, path information of abnormal structures, and corresponding view range for each type of result, and serves as the structural integrity verification result.

[0099] This step, based on the structural consistency verification results, further analyzes the integrity of conflicting local structures. By extracting the semantic subgraphs corresponding to the conflicting regions, a candidate subgraph set is constructed, which the system uses as the input set for structural enumeration analysis. A GTrie indexing mechanism is introduced, constructing an index tree with prefix-sharing characteristics from multiple standard template graphs in the structural pattern library, significantly improving the computational efficiency of graph structure matching. During the specific enumeration process, the method sequentially matches each candidate subgraph with the template structure in the GTrie, performing full-graph embedding verification based on node type, connection semantics, and topology to determine whether the model has a complete embedding of the structural template at the corresponding location. Finally, based on the matching results, the structural patterns are identified as normal, missing, redundant, or deformed items, forming a structural integrity verification result set, along with detailed information such as structural anomaly paths, view sources, and model element numbers, providing accurate basis for subsequent rule reasoning and defect localization. This process solves the problems of weak structural template recognition, ambiguous classification, and low processing efficiency in traditional model verification, significantly improving the comprehensiveness and engineering adaptability of structural integrity verification.

[0100] In this embodiment, S5 specifically includes:

[0101] S51. Import the structural consistency verification result, structural integrity verification result, and semantic attribute graph into the verification module as input, load the rule set, and define each rule as a rule structure composed of condition trigger body and response action;

[0102] S52. Initialize the matching of all model nodes in the semantic attribute graph using a rule-based matching initialization function. Construct a rule-based pattern matching index based on node attributes and edge relationships. Extract node combinations through matching patterns and generate conflict sets by combining defect information with the index structure. The conflict set generation function is defined as follows:

[0103] ;

[0104] in This represents the process by which the matching engine solves for the joint conditions between the rules and the model. This is a function for judging the consistency of semantic structure. For a set of rules, For semantic attribute graphs, For the k-th rule in the rule set, A collection of structural defect information;

[0105] S53. Traverse all rules that meet the conditions in the conflict set, add the matching model structure combination to the trigger queue, and perform structural relationship judgment, label verification and model state detection according to the rule action content. Extract the execution results and convert them into semantic feedback actions.

[0106] S54. Summarize the execution records of all triggered rules, and generate normative verification results based on semantic feedback actions. The results include the rule hit identifier, the corresponding model element number, the logical hit path, and the verification category marker.

[0107] This step, based on structural consistency, structural integrity results, and semantic attribute graphs, imports a rule set and constructs rule triggering conditions and response mechanisms. The system indexes model node attributes and connection relationships, performs matching initialization in conjunction with defect information, and forms a conflict set to locate potentially non-compliant structures. During the traversal of the conflict set, the system automatically identifies rules that meet the triggering conditions and performs logical judgments and semantic verification operations on the matched model structure combinations, extracting feedback information. Finally, the execution records of all triggered rules are uniformly summarized to generate standardized verification results containing the matched rule number, corresponding model element, trigger path, and verification category. This provides accurate and traceable structured evidence for model quality verification, significantly enhancing the intelligence and practicality of model verification.

[0108] In this embodiment, S6 specifically includes:

[0109] S61. Receive the structural consistency verification result, structural integrity verification result, and standardization verification result. Number the defect items contained in each type of verification result in a unified manner. Based on the model nodes and connection relationships in the semantic attribute graph, identify the model elements involved in each defect path, establish a mapping index table between defect number and model element number, and mark the dependency relationship between defect path and model structure.

[0110] S62. Based on the established mapping index table, analyze the node overlap between defect paths, divide defect items with structural overlap or connection coupling into the same structural cluster group, and each cluster group forms a structural verification result fragment, which includes the aggregated model element set, semantic path range, the model view to which it belongs and the trigger rule record, and generate a cluster group list.

[0111] S63. Perform structural mapping analysis on all cluster group lists, extract the distribution relationship of each group of model elements in the semantic attribute graph, identify cross-connection paths and merge duplicate structures, construct a structural influence map under multiple views, complete the dual alignment of defect clustering and view mapping, form a structural influence view and generate a unified representation format.

[0112] S64. Based on the path information, rule sources, structural influence areas and corresponding model element indexes in the structural influence view and model diagram, a unified structural verification report is generated. The report includes defect type classification, distribution statistics, cross-view distribution analysis, structural influence summary and rule trigger statistics.

[0113] This step assigns a unified number to defect items in the structural consistency, integrity, and standardization verification results, and establishes a mapping index between defect paths and model elements, forming defect structural dependencies. By analyzing the structural overlap and connection coupling of defect items in the model, the system divides interrelated defects into structural clusters, further constructing a structural impact map to complete the mapping alignment between defect clusters and view distribution. Finally, based on structural paths, rule sources, and element index information, the system uniformly generates a structural verification report, covering defect classification, cross-view distribution, impact structure analysis, and rule trigger statistics, providing structured and traceable data support for subsequent remediation decisions.

[0114] In this embodiment, S7 specifically includes:

[0115] S71. Using the structural verification report as input, construct a directed graph where nodes represent model elements and edges represent structural dependencies and defect propagation paths between elements. Label the initial fault source nodes based on the defect result set and form a defect propagation graph.

[0116] S72. Perform a minimum cut set search on the defect propagation graph to extract the minimum combination of nodes that covers all defect paths. The minimum cut set search expression is:

[0117] ;

[0118] in, For the set of minimal cut sets, , Both represent a single minimal cut set. To verify the set of nodes that affect graph a, For the first Defect path, This is a set of structural integrity verification results. This is a set of normative verification results. Representing cut set With defect path The intersection of the nodes is not an empty set. This indicates that there is no smaller cut set that can cover the same defect path;

[0119] S73. Calculate the total repair cost of each minimal cut set containing nodes, sort all minimal cut sets according to the cost function results, select the combination with the lowest cost as the priority repair candidate, and extract the model elements and their respective view and semantic path information contained in each group.

[0120] S74. Based on the sorted minimum cut set sequence, match the repair strategy template corresponding to the node type in the rule base, construct targeted repair suggestions, and output the defect location index, minimum cut set detailed mapping table, and structural repair suggestions.

[0121] This step constructs a defect propagation graph based on the structural verification report, using modeling elements as nodes and structural dependencies as edges to identify initial fault sources and trace defect paths. Subsequently, a minimum cut set search is performed on the propagation graph to extract the minimum combination of model elements covering all defect paths, forming the key area for defect localization. The system calculates and sorts the repair cost for each cut set, selecting the optimal repair candidate. Finally, based on the cut set results and matching the repair strategy templates in the rule base, a structured defect localization index, a minimum cut set mapping table, and repair suggestions are generated, improving the response efficiency and repair accuracy of design problems.

[0122] Example 1:

[0123] To verify the feasibility of this invention in practice, it was applied to the modeling, development, and verification process of a large-scale airborne communication and navigation system undertaken by an avionics system design and integration company. The problem addressed in this embodiment is that, in the traditional SysML modeling environment, there are significant information fragmentation and semantic disconnects between the company's multi-view models. Each team independently models the requirements view, structural view, and behavioral view, lacking effective methods for cross-view structural consistency and integrity analysis. This leads to delayed discovery of modeling conflicts, inability to promptly locate and correct design defects, resulting in repeated model modifications, extended design cycles, and increased project costs.

[0124] In practical applications, the company first designed the requirements diagram, internal structure diagram, and behavioral state diagram of the airborne communication and navigation system using the standard SysML modeling tool. The requirements diagram includes communication requirements and navigation accuracy requirements; the structure diagram defines the core communication controller, signal processing unit, navigation receiving module, and interface ports; and the behavioral diagram describes the key processes of the system, such as communication, signal processing, and navigation state switching. After completing the preliminary model design, the company exported the model in XMI format and conducted in-depth analysis using the model verification method proposed in this invention.

[0125] Using the model parsing module provided in this invention, the enterprise quickly and automatically extracted nearly 700 model elements from the exported SysML model file, including 200 requirement elements, 300 structural elements, and 200 behavioral elements, and automatically constructed a unified semantic attribute graph using a semantic mapping rule base. Next, the enterprise applied the VF2++ subgraph isomorphism algorithm to quickly identify multiple structural consistency defects in the model, including 10 areas where there are obvious conflicts between the requirement view and the structural view, and 15 areas where there are structural mismatches between the structural view and the behavioral view, totaling 25 structural consistency defects.

[0126] To address the structural consistency defects, the company further applied the GTrie graph pattern enumeration algorithm to perform structural integrity verification and structural template enumeration analysis on all identified defective areas. The verification identified 6 missing standard design structural templates, 3 redundant structural templates, and 4 semantically deformed structural templates, further clarifying the specific types and scope of the structural defects.

[0127] Subsequently, by applying the Rete-NT incremental rule engine of this invention, the enterprise achieved rapid and accurate model standardization verification, imported approximately 120 enterprise-customized model design rules, automatically performed intelligent rule reasoning and matching analysis on all possible defect structures in the semantic attribute graph, successfully hit 50 model node combinations that triggered the rule conditions, and generated a targeted standardization verification report.

[0128] After summarizing and analyzing all the above verification results, the company automatically generated a structured model verification report and used the minimum cut set analysis method to quickly locate four key defect propagation node sets. Combined with the design rule base, targeted defect remediation suggestions were generated. After designers quickly made repairs based on these suggestions, the number of model modifications and re-verifications decreased from an average of six to less than two, and the design and verification cycle was significantly shortened from 12 weeks to 6 weeks.

[0129] After implementing this invention, the company tracked and recorded data on efficiency improvements in similar airborne communication and navigation system design projects, resulting in detailed data as shown in Table 1:

[0130] Table 1. Data on the Implementation Results of the Model Verification of the Avionics Airborne Communication and Navigation System

[0131] Project Indicators Before implementing this invention After implementing this invention Improvement rate (%) Total number of model elements (number) 700 700 - Number of structural defects found across views (sites) 10 25 +150% Model structure template defect identification accuracy (%) 65% 95% +46.15% Average number of hits in rule-based reasoning (points) 20 50 +150% Average model defect localization accuracy (%) 55% 92% +67.27% Average number of repair cycles (times) 6 2 -66.67% Average total cycle time for modeling and validation (weeks) 12 6 -50.00% Total project cost (ten thousand yuan) 90 58 -35.56%

[0132] As shown in Table 1, by implementing the model verification method proposed in this invention, the number of cross-view structural defects discovered has increased significantly by 150%, the accuracy of structural template defect identification has increased from 65% to 95%, the accuracy of model defect location has increased by 67.27%, the average number of repair cycles has decreased by two-thirds, the overall modeling and verification cycle has been shortened by 50%, and the overall project cost has been reduced by 35.56%. This demonstrates that the invention has significant beneficial effects in improving model verification accuracy, improving design quality, and saving project time and costs, and fully meets the expected purpose and requirements of implementing the invention.

[0133] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A model validation method based on SysML language, characterized in that, Includes the following steps: S1. Obtain the SysML model file to be verified and parse it to extract model elements and metadata, and generate a set of model elements containing multi-view element information. S2. Input the set of model elements into the semantic modeling module and construct a unified semantic attribute graph according to the preset semantic mapping rules; S3. Using the semantic attribute graph as input, apply the VF2++ subgraph isomorphism algorithm to perform matching analysis on the structural subgraphs in the semantic attribute graph, identify structural consistency conflicts or missing relationships between cross-views, and output the structural consistency verification results. S4. Based on the structural consistency verification result, the GTrie graph pattern enumeration algorithm is further called to perform structural pattern recognition on the node combination to determine whether there is a missing, deformed or redundant design structural template, and output the structural integrity verification result. S5. Take the structural integrity verification result and the semantic attribute graph as joint input, load the rule base and start the Rete-NT incremental rule engine, perform verification according to the structural rules, behavioral rules and user-defined rules in the rule base, and output the normative verification result. S6. Summarize the results of structural consistency verification, structural integrity verification, and conformity verification to generate a unified structural verification report; S7. Taking the structural verification report as input, execute the minimum cut set generation algorithm to determine the minimum set of model elements that cause the defect based on the dependency path between defective elements, and generate corresponding repair suggestions according to the rule base, and output a complete defect location and repair plan.

2. The model validation method based on SysML language according to claim 1, characterized in that, S1 specifically includes: S11. Load the SysML model file to be verified, identify the storage format of the SysML model file, and read the model file hierarchically through a parser based on the document object model and the event-driven model. Identify the package structure, model segments and the starting position of various views in the model, and provide an index structure for model element extraction. S12. Based on the index structure, extract model elements and three types of model view content from the model file as needed, and simultaneously record the level of the model view content, parent-child relationship, and cross-graph reference node path. S13. Call the metadata extraction module to read the metadata information of each model element; S14. The extracted model elements and corresponding metadata are structurally encapsulated and divided into model element sets. Integrity checks and duplicate identification filtering are performed on the elements in the model element sets to remove missing attributes and illegal nodes, and finally a model element set in a unified format is generated.

3. The model validation method based on SysML language according to claim 1, characterized in that, S2 specifically includes: S21. Based on the received model element set, call the semantic mapping rule library to parse the attribute fields and relation fields of the model element set, and map the different types of requirement elements, structural elements and behavioral elements to a unified data representation model according to the semantic mapping rules. S22. Represent each model element as a node in a graph structure, and establish semantic connection edges between nodes according to the various relationships between model elements to form an attribute graph containing a set of nodes and a set of edges. S23. Perform structural optimization processing on the completed attribute graph, supplement the view source type, model semantic label and cross-view matching identifier of each node, add semantic weight information to all edges, and form a semantic attribute graph.

4. The model validation method based on SysML language according to claim 1, characterized in that, S3 specifically includes: S31. Using the constructed semantic attribute graph as the input graph, read the structural information of all nodes and edges in the graph, and extract the view source, type label and naming attribute of each node, as well as the semantic identifier and direction information of each edge. S32. Load a target structure subgraph from the structure pattern library as a reference template for consistency verification. Each node in the target structure subgraph represents a model component and has a type label and semantic connection requirements, forming a standardized structure template. S33. In the semantic attribute graph, call the VF2++ subgraph isomorphism algorithm to perform structure matching on the standardized structure template, construct a mapping function from the template subgraph to the input graph, and satisfy the mapping verification condition. The expression for the mapping verification condition is: ; in, , This represents two connected nodes in the template diagram. This represents the mapping result of nodes in the semantic attribute graph. The set of edges representing the semantic attribute graph. Indicates the template node type label, Indicates a node type tagging function, Semantic labels representing edges in a structural template graph; S34. Identify the nodes and edges that can pass the mapping verification conditions, output the structural consistency verification results, and locate the nodes and edges that fail the mapping verification conditions as areas with inconsistent structures, and record the view type, node name, edge connection type and upstream and downstream relationship.

5. The model validation method based on SysML language according to claim 1, characterized in that, S4 specifically includes: S41. Based on the structural consistency verification results, extract all regions with structural conflicts in the semantic attribute graph, locate the subgraph range corresponding to each conflict region, and extract each range as a local subgraph containing a set of nodes and a set of edges, and form a candidate subgraph set from all local subgraphs. S42. Load multiple structural template graphs from the structural pattern library, construct a GTrie index tree by sharing all structural template graphs through graph prefixes based on the candidate subgraph set, establish the index mapping relationship between node topology paths and structural patterns, and form a set of structural template indexes for enumeration and traversal. S43. For each subgraph in the candidate subgraph set, call the target structure template graph from the GTrie structure in turn, perform the structure template enumeration operation in the current subgraph by graph traversal, and define the structure pattern matching function. S44. Based on the graph structure enumeration matching results, record the templates that are successfully matched, the templates that are not successfully matched, the template instances whose structures in the candidate subgraphs exceed the template definition range, and the template instances whose semantic connections and node types are inconsistent, and integrate them into the structural integrity verification results.

6. The model validation method based on SysML language according to claim 1, characterized in that, S5 specifically includes: S51. Import the structural consistency verification result, structural integrity verification result, and semantic attribute graph into the verification module as input, load the rule set, and define each rule as a rule structure composed of condition trigger body and response action; S52. Initialize the matching of all model nodes in the semantic attribute graph using a rule-based matching initialization function. Construct a rule-based pattern matching index based on node attributes and edge relationships. Extract node combinations through matching patterns and generate conflict sets by combining defect information with the index structure. The conflict set generation function is defined as follows: ; in This represents the process by which the matching engine solves for the joint conditions between the rules and the model. This is a function for judging the consistency of semantic structure. For a set of rules, For semantic attribute graphs, For the k-th rule in the rule set, A collection of structural defect information; S53. Traverse all rules that meet the conditions in the conflict set, add the matching model structure combination to the trigger queue, and perform structural relationship judgment, label verification and model state detection according to the rule action content. Extract the execution results and convert them into semantic feedback actions. 7.S54. Summarize the execution records of all triggered rules and generate normative verification results based on semantic feedback actions.

8. The model validation method based on SysML language according to claim 1, characterized in that, S6 specifically includes: S61. Receive the structural consistency verification result, structural integrity verification result, and standardization verification result. Number the defect items contained in each type of verification result in a unified manner. Based on the model nodes and connection relationships in the semantic attribute graph, identify the model elements involved in each defect path, establish a mapping index table between defect number and model element number, and mark the dependency relationship between defect path and model structure. S62. Based on the established mapping index table, analyze the node overlap between defect paths, divide defect items with structural overlap or connection coupling into the same structural cluster group, and generate a list of cluster groups. S63. Perform structural mapping analysis on all cluster group lists, extract the distribution relationship of each group of model elements in the semantic attribute graph, identify cross-connection paths and merge duplicate structures, construct a structural influence map under multiple views, complete the dual alignment of defect clustering and view mapping, and form a structural influence view. S64. Based on the path information, rule source, structural influence area and corresponding model element index in the structural influence view and model diagram, generate a unified structural verification report.

9. The model validation method based on SysML language according to claim 1, characterized in that, Specifically, S7 includes: S71. Using the structural verification report as input, construct a directed graph, label the initial fault source nodes according to the defect result set, and form a defect propagation graph; S72. Perform a minimum cut set search on the defect propagation graph to extract the minimum combination of nodes that covers all defect paths. The minimum cut set search expression is: ; in, For the set of minimal cut sets, , Both represent a single minimal cut set. To verify the set of nodes that affect graph a, For the first Defect path, This is a set of structural integrity verification results. This is a set of normative verification results. Representing cut set With defect path The intersection of the nodes is not an empty set. This indicates that there is no smaller cut set that can cover the same defect path; S73. Calculate the total repair cost of each minimal cut set containing nodes, sort all minimal cut sets according to the cost function results, select the combination with the lowest cost as the priority repair candidate, and extract the model elements and their respective view and semantic path information contained in each group. S74. Based on the sorted minimum cut set sequence, match the repair strategy template corresponding to the node type in the rule base, construct targeted repair suggestions, and output the defect location index, minimum cut set detailed mapping table, and structural repair suggestions.