CN-121638475-B - MBSE-oriented multi-agent collaborative automation modeling method and system

CN121638475BCN 121638475 BCN121638475 BCN 121638475BCN-121638475-B

Abstract

The invention relates to the technical field of model driven system engineering modeling, in particular to a MBSE-oriented multi-agent collaborative automatic modeling method and system. The method comprises the steps of obtaining natural language modeling requirements of users, retrieving knowledge subgraphs related to the requirements in a shared knowledge environment, constructing modeling context, calling and generating an agent to generate a current SysML v2 model, calling and generating the agent to perform grammar verification and semantic verification based on domain knowledge graphs, outputting verification results, calling and restoring the agent to execute deterministic restoration and iterative verification according to the domain knowledge graphs when the verification result representation grammar fails or has semantic inconsistency or violates engineering constraint rules, and outputting the final SysML v2 model under the condition of meeting preset conditions, wherein notch marking or bottom-covering restoration can be performed under the condition of covering notches. The invention improves the accuracy, the interpretability and the convergence efficiency of complex system modeling.

Inventors

DI XIAOQIANG
QI HUI
CAO JINHUI
LI ZHI
PENG KUN
LI SHIHAO
HE XIONGWEN
LI JINQING
ZHOU SHIYING
YANG HUAMIN
LIU XU

Assignees

长春理工大学

Dates

Publication Date: 20260508
Application Date: 20260204

Claims (10)

1. MBSE-oriented multi-agent collaborative automation modeling method is characterized by comprising the following steps of: S1, acquiring a natural language modeling requirement input by a user; Step S2, based on the natural language modeling requirement, retrieving a knowledge sub-graph related to the natural language modeling requirement in a shared knowledge environment, and constructing a modeling context based on the knowledge sub-graph, wherein the shared knowledge environment at least comprises a domain knowledge graph, and the domain knowledge graph is used for representing a SysML v2 meta-model, a domain ontology and engineering constraint rules; S3, calling to generate an agent, and generating a current SysML v2 model according to the modeling context; s4, invoking a verification intelligent agent, carrying out grammar verification on the current SysML v2 model and semantic verification based on the domain knowledge graph, and outputting a verification result; And S5, when the verification result represents that the grammar verification of the current SysML v2 model fails or semantic inconsistency exists or the engineering constraint rule is violated, invoking a repair agent to execute deterministic repair according to the domain knowledge graph to obtain a repaired SysML v2 model, and returning to the step S4 until the verification result represents that the repaired SysML v2 model meets the preset condition, and outputting a final SysML v2 model.
2. The MBSE-oriented multi-agent collaborative automation modeling method according to claim 1, wherein in step S2: performing graph retrieval enhancement in the shared knowledge environment based on the natural language modeling requirements to obtain the knowledge subgraph; The graph retrieval enhancement comprises determining retrieval elements according to the natural language modeling requirement, matching the retrieval elements in the domain knowledge graph to obtain seed nodes, and performing Jump neighborhood expansion to form a local connected subgraph related to the natural language modeling requirement as the knowledge subgraph; The knowledge subgraph is serialized to build the modeling context.
3. The method for collaborative automation modeling of multiple agents for MBSE in accordance with claim 2, wherein the shared knowledge environment further includes a vector database for storing semantic vector representations of nodes and/or relationships in the domain knowledge graph, the graph retrieval enhancement includes vectorizing the natural language modeling requirements, performing similarity retrieval based on the semantic vectors of the natural language modeling requirements and the semantic vectors in the vector database to obtain candidate seed nodes, and performing neighborhood expansion in the domain knowledge graph with the candidate seed nodes to obtain the knowledge subgraph.
4. The MBSE-oriented multi-agent collaborative automation modeling method according to claim 1, wherein the verification result at least comprises a semantic decision result of the current SysML v2 model, the semantic decision result is one of a strict valid state, a weak valid state and an invalid state, wherein the strict valid state is used for representing that grammar verification is passed and the semantic verification does not detect conflict violating the engineering constraint rule, the weak valid state is used for representing that grammar verification is passed and the semantic verification does not detect conflict but a domain knowledge graph coverage gap exists, and the invalid state is used for representing that grammar verification is not passed or the semantic verification detects conflict violating the engineering constraint rule.
5. The MBSE-oriented multi-agent collaborative automation modeling method according to claim 1, wherein the semantic verification includes converting the current SysML v2 model into a temporary model graph, nodes of the temporary model graph are used for representing system constituent elements and interfaces or ports thereof, edges of the temporary model graph are used for representing interface connection relations, constraint consistency check is performed based on the temporary model graph and the domain knowledge graph, and the constraint consistency check includes judging sub-graph matching relations and checking engineering constraint rules to output the verification result.
6. The MBSE-oriented multi-agent collaborative automation modeling method according to claim 1, wherein the grammar validation includes parsing the current SysML v2 model based on SysML v2 grammar rules to determine whether the current SysML v2 model satisfies a preset grammar constraint and outputting grammar error information when not satisfied, the grammar error information includes at least an error location and an error type, and the deterministic repair in step S5 includes grammar correction of the current SysML v2 model according to the grammar error information.
7. The MBSE-oriented multi-agent collaborative automation modeling method of claim 4, wherein when the semantic decision is weak and the validation result characterizes a domain knowledge graph coverage gap, the method further comprises at least one of: Performing overlay notch marking on the current SysML v2 model, and synchronously outputting the overlay notch marking when outputting the final SysML v2 model; And calling the repairing agent to execute spam repairing to update the current SysML v2 model, and returning to the step S4 to verify again, wherein the spam repairing comprises complementing or rewriting the relevant part of the coverage gap based on a large language model.
8. The MBSE-oriented multi-agent collaborative automation modeling method according to claim 4, wherein the deterministic repair in step S5 includes generating a structured error report based on the verification result, determining an error type according to the structured error report, retrieving a repair path or an alternative structure satisfying the engineering constraint rule in the domain knowledge graph for the error type, and performing a corresponding repair operation on the current SysML v2 model according to the retrieval result to obtain the post-repair SysML v2 model.
9. The MBSE-oriented multi-agent collaborative automation modeling method of claim 8, wherein the preset conditions include at least one of: the semantic judgment result obtained in the step S4 is strictly effective; The semantic judgment result obtained in the step S4 is weak and effective, a coverage gap mark is generated, and the coverage gap mark is associated with the final SysML v2 model and output; The iteration number of the deterministic repair reaches a preset upper limit, or the structured error report is not changed in iteration of continuous preset times, so that iteration is terminated and the current SysML v2 model is output as the final SysML v2 model.
10. MBSE-oriented multi-agent collaborative automation modeling system, characterized in that the system comprises: the demand acquisition module is used for acquiring natural language modeling demands input by a user; the sub-graph searching and context constructing module is used for searching and obtaining a knowledge sub-graph related to the natural language modeling requirement in a shared knowledge environment based on the natural language modeling requirement and constructing a modeling context based on the knowledge sub-graph, wherein the shared knowledge environment at least comprises a domain knowledge graph, and the domain knowledge graph is used for representing a SysML v2 meta-model, a domain ontology and engineering constraint rules; the model generation module is used for calling and generating an agent and generating a current SysML v2 model according to the modeling context; The model verification module is used for calling a verification intelligent agent, carrying out grammar verification on the current SysML v2 model and semantic verification based on the domain knowledge graph, and outputting a verification result; And the model repair and output module is used for calling a repair intelligent body to execute deterministic repair according to the domain knowledge graph to obtain a repaired SysML v2 model when the verification result represents that the grammar verification of the current SysML v2 model fails or semantic inconsistency exists or the engineering constraint rule is violated, returning the repaired SysML v2 model to the model verification module for re-verification until the verification result represents that the repaired SysML v2 model meets the preset condition, and outputting a final SysML v2 model.

Description

MBSE-oriented multi-agent collaborative automation modeling method and system Technical Field The invention relates to the technical field of model driven system engineering modeling, in particular to a MBSE-oriented multi-agent collaborative automatic modeling system. Background In the process of gradually replacing traditional document type system engineering by model driven system engineering (MBSE), the system model is used as a formalized carrier throughout the whole life cycle for demand analysis, architecture design, verification validation and collaborative delivery. Around this paradigm, sysML has evolved continuously since 2007 was adopted by OMG, which adopted SysML v1.7 at 2024, month 6, and next generation SysML v2.0 at 2025, month 6, laying a standard foundation for accurate expression, interoperation, and automation of models. In recent years, the natural language understanding and code generating capability of a large language model enables automatic modeling from natural language to SysML to be a reality direction, related research gradually evolves from early relying on static constraint of a structural template to generation closed loop combining retrieval enhancement and a parser, and further goes to an automatic modeling paradigm of multi-agent cooperation so as to reduce modeling threshold and improve modeling efficiency. The existing automatic modeling technology based on a large language model can improve the generation efficiency, but still faces the general problem of 'grammar correct but engineering semantic invalid' in a complex system scene, namely, model text can be resolved, but hidden errors exist in interface matching, constraint consistency and topology validity, so that the risk of uncontrollable engineering is caused. On the one hand, the traditional Text RAG mainly recalls Text fragments by vector similarity, and is difficult to keep system topology and constraint information, and the unstructured context can amplify semantic illusion risks, so that connection relation and constraint inference are unreliable. On the other hand, many closed-loop repairing processes still highly depend on probability trial and error of a large model, so that the phenomena of high iteration turn, unstable convergence, new error introduction during repairing an error and the like are easy to occur, and particularly, the phenomena of difficult stable correction during semantic violations related to field constraint or cross-element relation are more difficult. Meanwhile, when the knowledge in the field is incomplete, the system can only be degraded and depends on a weak effective strategy, so that the determinability and the interpretability are further reduced. Therefore, a MBSE-oriented multi-agent collaborative automation modeling method and system are needed to solve the problems. Disclosure of Invention (One) solving the technical problems Aiming at the defects of the prior art, the invention provides MBSE-oriented multi-agent collaborative automatic modeling method and system, which solve the problems. (II) technical scheme In order to achieve the purpose, the invention provides a MBSE-oriented multi-agent collaborative automatic modeling method, which comprises the following steps of S1, acquiring natural language modeling requirements input by a user. Step S2, based on the natural language modeling requirement, retrieving a knowledge sub-graph related to the natural language modeling requirement in a shared knowledge environment, and constructing a modeling context based on the knowledge sub-graph, wherein the shared knowledge environment at least comprises a domain knowledge graph, and the domain knowledge graph is used for representing a SysML v2 meta-model, a domain ontology and engineering constraint rules; And S3, calling to generate an agent, and generating a current SysML v2 model according to the modeling context. And S4, calling a verification agent, carrying out grammar verification on the current SysML v2 model and semantic verification based on the domain knowledge graph, and outputting a verification result. And S5, when the verification result represents that the grammar verification of the current SysML v2 model fails or semantic inconsistency exists or the engineering constraint rule is violated, invoking a repair agent to execute deterministic repair according to the domain knowledge graph to obtain a repaired SysML v2 model, and returning to the step S4 until the verification result represents that the repaired SysML v2 model meets the preset condition, and outputting a final SysML v2 model. Further, in the step S2, graph retrieval enhancement is performed in the shared knowledge environment based on the natural language modeling requirements to obtain the knowledge subgraph. The graph retrieval enhancement comprises determining retrieval elements according to the natural language modeling requirement, matching the retrieval elements in the domain knowledge graph to