Fusion method and device of multi-source heterogeneous model, storage medium and electronic equipment

Abstract

The invention discloses a multi-source heterogeneous model fusion method and device, a storage medium and electronic equipment, and belongs to the technical field of computer intelligence. The method comprises the following steps: acquiring a multivariate heterogeneous sub-model to be fused; analyzing the sub-models to obtain personalized features of the multivariate heterogeneous sub-models to be fused, and constructing a unified semantic representation system based on a meta-model framework; according to the personalized features of the sub-models, standardized expression is carried out on the logic expressions of the sub-models by utilizing a unified intermediate expression language to obtain a to-be-fused multivariate heterogeneous sub-model with a standardized expression form; and according to the sub-model with the standardized expression form, a multivariate heterogeneous model based on the to-be-fused multivariate heterogeneous sub-model is constructed and obtained based on the knowledge graph and the graph neural network. The device, the storage medium and the electronic equipment can be used for implementing the method, and the efficiency and reliability of multi-model collaborative design and simulation verification in the development process of complex equipment products can be remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of computer intelligence technology, and in particular to a method, device, storage medium and electronic equipment for fusing multi-source heterogeneous models. Background Art

[0002] In the existing technology, the following mainstream technical routes and their limitations exist in the field of multi-source heterogeneous model integration: Model-driven architecture achieves basic model integration through a conversion mechanism from platform-independent models to platform-specific models. Although its model conversion method based on abstraction levels has advantages in terms of architectural separation, it essentially relies on syntactic mapping between metamodels and lacks in-depth analysis and association of semantic information such as domain knowledge and business rules, making it difficult to achieve accurate semantic consistency in complex industrial scenarios. Ontology-based conversion technology (such as using OWL ontology to construct the mapping relationship between SysML requirement models and Simulink design models) has made some breakthroughs in semantic enhancement, but its conversion rules are often fixed in a predefined static ontology framework. When faced with scenarios such as dynamically changing engineering constraints and real-time updated design parameters, they have problems such as insufficient rule adaptability and high maintenance costs. Data middleware technology (such as XML / JSON format conversion) is a lightweight solution. Although it achieves surface data interoperability through standardized data formats, it is easy to lose key semantic information such as topological relationships and behavioral logic between models during the conversion process, making cross-model data lineage tracing difficult and unable to ensure logical consistency during multi-model collaborative simulation. Summary of the Invention

[0003] In view of this, the present invention provides a method, device, storage medium and electronic device for fusing multi-source heterogeneous models, which can significantly improve the efficiency and reliability of multi-model collaborative design and simulation verification in the development process of complex equipment products, making it more suitable for practical use.

[0004] In order to achieve the first objective above, the technical solution of the multi-source heterogeneous model fusion method provided by the present invention is as follows:

[0005] The multi-source heterogeneous model fusion method provided by the present invention comprises the following steps:

[0006] Obtaining the multivariate heterogeneous sub-models to be fused;

[0007] Analyze the multi-heterogeneous sub-models to be fused, obtain personalized features of the multi-heterogeneous sub-models to be fused, and construct a unified semantic representation system based on the metamodel framework;

[0008] According to the personalized characteristics of the multi-heterogeneous sub-models to be fused, the logical expressions of the multi-heterogeneous sub-models to be fused are standardized using a unified intermediate representation language to obtain the multi-heterogeneous sub-models to be fused with a standardized expression form;

[0009] According to the multivariate heterogeneous sub-models to be fused with standardized expressions, a multivariate heterogeneous model based on the multivariate heterogeneous sub-models to be fused is constructed based on the knowledge graph and graph neural network.

[0010] The multi-source heterogeneous model fusion method provided by the present invention can also be further implemented by adopting the following technical measures.

[0011] Preferably, the method for fusing multi-source heterogeneous models further comprises the following steps:

[0012] Based on the three-dimensional interactive visualization interface, a visualization engine and multi-dimensional verification system are constructed, including:

[0013] The 3D interactive visualization interface displays model node topology, data flow paths, and conflict warnings;

[0014] The verification system includes static verification and dynamic verification, and supports model version backtracking and evolution deduction.

[0015] Preferably, the personalized features of the multi-heterogeneous sub-models to be integrated include the logical rules of the multi-heterogeneous sub-models, the constraints of the demand model, the topological structure of the architecture model, the parameterized features of the design model, and the manufacturing knowledge of the process model.

[0016] Preferably, parsing the multivariate heterogeneous sub-models to be fused to obtain personalized features of the multivariate heterogeneous sub-models to be fused, and constructing a unified semantic representation system based on a metamodel framework specifically includes the following steps:

[0017] The model parsing engine uses differentiated parsing strategies for different model types. It extracts the functional chain and logic rules of SysML models and identifies concurrent execution logic in activity diagrams through semantic analysis. It also performs structured parsing of requirement models and uses natural language processing to extract requirement items and their constraints. It uses the Modelica interface to parse the topological connections and physical equations of the architecture model and extract state variables and parameter dependencies in differential algebraic equations. It uses the CAD software API to extract the geometric features, parametric constraints, and assembly relationships of the design model. Furthermore, it extracts manufacturing processes, equipment parameters, and process constraints from the process knowledge base to construct manufacturing feasibility assessment indicators for the process model.

[0018] Constructing a multidimensional meta-model framework based on ontology language to uniformly represent the core attributes of heterogeneous models, wherein the dimensions of the ontology language include functional dimension, structural dimension, behavioral dimension, domain semantics, and interface specification level;

[0019] The parsed model metadata is converted into standardized RDF triples and stored in the Neo4j graph database to form a unified data foundation that supports cross-model fusion and conversion.

[0020] Preferably, according to the multivariate heterogeneous sub-models to be fused having standardized expressions, constructing a multivariate heterogeneous model based on the multivariate heterogeneous sub-models to be fused based on a knowledge graph and a graph neural network comprises the following steps:

[0021] Differentiated parsing algorithms are used for different model types. The BERT+BiLSTM model is used to perform intent recognition and context analysis on the requirement text, extract the requirement constraint vector, and generate function labels.

[0022] Graph convolutional networks are used to learn the graph structure of SysML activity diagrams, capturing the timing dependencies and parallel execution characteristics in functional logic. Tensor decomposition technology is used to reduce the dimensionality of the topological connection relationships in the architecture model and encode them into a low-dimensional dense matrix to represent the structural characteristics.

[0023] Construct a five-dimensional feature matrix covering functions, structures, behaviors, domain semantics, and interface specifications, where the feature vector of each model instance contains functional labels, structural features, dynamic behavior attributes, domain semantic indicators, and interface compatibility parameters;

[0024] Cross-model association analysis is achieved through a semantic similarity calculation model. A weighted fusion strategy is adopted to integrate the cosine similarity of functional features, the Jaccard similarity of structural features, and the KL divergence of domain semantics to generate the semantic association strength value between models. The weight coefficients (α, β, γ) are dynamically optimized through a genetic algorithm to adapt to the model fusion conversion and coordination requirements in different application scenarios, ultimately forming a cross-model semantic association network that supports fusion conversion.

[0025] Preferably, according to the multivariate heterogeneous sub-models to be fused in a standardized expression form, based on the knowledge graph and the graph neural network, constructing a multivariate heterogeneous model based on the multivariate heterogeneous sub-models to be fused further includes the following steps:

[0026] Through the intermediate semantic layer, parameter mapping function library and interface adaptation engine, dynamic semantic alignment and interface interoperability between multi-source heterogeneous models are achieved. Timestamp synchronization and dynamic buffer scheduling are used to ensure the timing consistency of data interaction, realizing plug-and-play across model interfaces.

[0027] Preferably, according to the multivariate heterogeneous sub-models to be fused in a standardized expression form, based on the knowledge graph and the graph neural network, constructing a multivariate heterogeneous model based on the multivariate heterogeneous sub-models to be fused further includes the following steps:

[0028] With the system model as the root node, the requirement model as the constraint node, the architecture model as the structure node, the design model as the leaf node, and the process model as the end node, a hierarchical knowledge graph is constructed, and four types of associations are defined: requirement coverage edge, structure mapping edge, process constraint edge, and dynamic evolution edge.

[0029] The GraphSAGE algorithm is used to embed graphs and generate low-dimensional vectors to represent the functional coupling, structural dependency and semantic similarity between models.

[0030] In order to achieve the second objective above, the technical solution of the multi-source heterogeneous model fusion method and device provided by the present invention is as follows:

[0031] The fusion device of the multivariate heterogeneous model provided by the present invention includes:

[0032] Multi-source heterogeneous sub-model acquisition module, used to obtain multi-source heterogeneous sub-models to be fused;

[0033] A unified semantic representation system construction module is used to parse the multi-heterogeneous sub-models to be fused, obtain personalized features of the multi-heterogeneous sub-models to be fused, and construct a unified semantic representation system based on the metamodel framework;

[0034] A standardized expression module is used to standardize the logical expressions of the multivariate heterogeneous sub-models to be fused according to the personalized characteristics of the multivariate heterogeneous sub-models to be fused using a unified intermediate representation language to obtain the multivariate heterogeneous sub-models to be fused with a standardized expression form;

[0035] A multivariate heterogeneous model construction module is used to construct a multivariate heterogeneous model based on the multivariate heterogeneous sub-models to be fused based on the knowledge graph and graph neural network according to the multivariate heterogeneous sub-models to be fused with standardized expressions.

[0036] In order to achieve the third objective above, the technical solution of the computer-readable storage medium provided by the present invention is as follows:

[0037] The computer-readable storage medium provided by the present invention stores a fusion program of multi-source heterogeneous models. When the fusion program of multi-source heterogeneous models is executed by a processor, the steps of the fusion method of multi-source heterogeneous models provided by the present invention are implemented.

[0038] In order to achieve the fourth objective, the present invention provides an electronic device with the following technical solutions:

[0039] The electronic device provided by the present invention includes a memory and a storage, wherein the memory stores a fusion program of multi-source heterogeneous models. When the fusion program of multi-source heterogeneous models is executed by the processor, the steps of the fusion method of multi-source heterogeneous models provided by the present invention are implemented.

[0040] The present invention provides a method, device, and electronic device for fusing multiple heterogeneous models, offering efficient collaboration and dynamic adaptation capabilities, as well as an innovative mechanism for dynamic cross-model conversion. Specifically, by constructing an automated semantic alignment mechanism based on a multi-level metamodel, the present invention achieves efficient collaboration among multiple heterogeneous models, including system models, requirement models, architecture models, design models, and process models. This mechanism effectively addresses the low inter-model collaboration efficiency and poor dynamic adaptability inherent in traditional integration methods. Furthermore, the use of a dynamic rule generation engine addresses dynamic scenarios such as requirement changes and parameter iterations, addressing the limitations of traditional integration methods that rely on static mapping rules. Regarding dynamic cross-model conversion, the present invention utilizes an intermediate semantic layer, a parameter mapping function library, and an interface adaptation engine to achieve dynamic semantic alignment and interface interoperability among multiple heterogeneous models. The intermediate semantic layer defines a unified intermediate representation language, converting the logical expressions of different models into a standardized form. Developed based on middleware, the interface adaptation engine supports real-time conversion across multiple protocols and formats, and ensures the timing consistency of data interaction through timestamp synchronization and dynamic buffer scheduling. This design not only realizes plug-and-play of cross-model interfaces, but also ensures the efficiency and reliability of data interaction, greatly improving the overall performance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present invention. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0042] Attachment Figure 1 A flowchart of the steps of the fusion method of multi-source heterogeneous models provided by an embodiment of the present invention;

[0043] Attachment Figure 2 A schematic diagram of the signal flow relationship between the functional modules in the multi-source heterogeneous model fusion device provided by an embodiment of the present invention;

[0044] Attachment Figure 3 A schematic diagram of the device structure of a multi-source heterogeneous model fusion method for a hardware operating environment provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to solve the problems existing in the prior art, the present invention provides a method, device, storage medium and electronic equipment for fusing multi-source heterogeneous models, which can significantly improve the efficiency and reliability of multi-model collaborative design and simulation verification in the development process of complex equipment products, thereby being more suitable for practical use.

[0046] Glossary:

[0047] Multi-source heterogeneous model: refers to a model built using different modeling languages ​​(such as SysML, UML, Modelica), different abstraction levels (system level, requirement level, architecture level, design level) and different tools (such as MagicDraw, Matlab, MWorks), with semantic and structural differences.

[0048] System model: A model that describes the overall structure, functional interactions, and external environment of a system, usually built based on the DoDAF or UPDM framework.

[0049] Requirements Model: In model-based systems engineering, a model used to express system requirements. It can fully or partially express requirements at any level of abstraction, typically describing the system's functionality, attributes, and quality factors in a structured manner.

[0050] Architectural model: A model used in the systems engineering process to define, analyze, and verify the system architecture. It describes the system's components, interfaces, data flows, and interactions in a structured manner, providing a unified framework for system development and maintenance.

[0051] Semantic mapping: Establish logical associations between elements of different models, for example, mapping the "response time ≤ 10ms" of the requirement model to the "processor main frequency ≥ 2GHz" of the design model.

[0052] Model transformation rules: define the logic for transforming elements from the source model to the target model, including syntax transformation (such as XML to JSON) and semantic transformation.

[0053] After arduous efforts, the inventor discovered that:

[0054] (1) In the field of multi-source heterogeneous model integration, the current mainstream technology routes have the following limitations:

[0055] Model-driven architecture: This architecture achieves model integration at a basic level through a conversion mechanism from platform-independent models to platform-specific models. Although its model conversion method based on abstraction levels has advantages in terms of architectural separation, it essentially relies on grammatical mapping between metamodels and lacks in-depth analysis and association of semantic information such as domain knowledge and business rules. This makes it difficult to achieve accurate semantic consistency assurance in complex industrial scenarios.

[0056] Ontology-based conversion technology: For example, using OWL ontology to construct the mapping relationship between SysML requirement models and Simulink design models. Although this technology has made some breakthroughs in semantic enhancement, its conversion rules are often solidified in a predefined static ontology framework. When faced with scenarios such as dynamically changing engineering constraints and real-time updated design parameters, it suffers from problems such as insufficient rule adaptability and high maintenance costs.

[0057] Data middleware technology: such as XML / JSON format conversion, as a lightweight solution, although it achieves surface data interoperability through standardized data formats, it is easy to lose key semantic information such as topological relationships and behavioral logic between models during the conversion process, making it difficult to trace data lineage across models and unable to ensure logical consistency during multi-model collaborative simulation.

[0058] (2) In response to the shortcomings of the existing technology, the present invention proposes innovative solutions, the core objectives of which include:

[0059] Automated semantic alignment mechanism based on multi-level metamodel: By integrating domain ontology libraries and cross-model constraint rules, an extensible unified model semantic graph is constructed to achieve semantic ambiguity elimination and bidirectional reasoning of parameter mapping relationships between system models, requirement models, architecture models and structural design models, supporting intelligent matching and conversion of multi-domain models throughout the entire life cycle of model-based system engineering.

[0060] An automated conversion mapping mechanism between heterogeneous models is built based on a dynamic rule generation engine: it supports dynamic and flexible configuration of fusion conversion strategies and can achieve incremental updates of mapping relationships along with the model iteration process.

[0061] To further illustrate the technical means and effectiveness of the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the multi-source heterogeneous model fusion method, apparatus, storage medium, and electronic device proposed in accordance with the present invention, including their specific implementation, structure, features, and effectiveness. In the following description, different references to "one embodiment" or "embodiment" do not necessarily refer to the same embodiment. Furthermore, features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0062] The term "and / or" in this article is merely a description of the association relationship between associated objects, indicating that three relationships may exist. For example, A and / or B, specifically understood as: A and B may be included at the same time, A may exist alone, or B may exist alone, and any of the above three situations may exist.

[0063] Fusion method of multi-source heterogeneous models

[0064] See attached Figure 1The multi-source heterogeneous model fusion method provided by the embodiment of the present invention includes the following steps:

[0065] Step S1: obtaining the multivariate heterogeneous sub-model to be fused;

[0066] Step S2: Analyze the multi-heterogeneous sub-models to be fused, obtain the personalized features of the multi-heterogeneous sub-models to be fused, and construct a unified semantic representation system based on the metamodel framework;

[0067] Step S3: Based on the personalized characteristics of the multivariate heterogeneous sub-models to be fused, the logical expressions of the multivariate heterogeneous sub-models to be fused are standardized using a unified intermediate representation language to obtain the multivariate heterogeneous sub-models to be fused with a standardized expression form;

[0068] Step S4: Based on the multivariate heterogeneous sub-models to be fused with standardized expressions, a multivariate heterogeneous model based on the multivariate heterogeneous sub-models to be fused is constructed based on the knowledge graph and graph neural network.

[0069] The fusion method of multi-source heterogeneous models also includes the following steps:

[0070] Based on the three-dimensional interactive visualization interface, a visualization engine and multi-dimensional verification system are constructed, including:

[0071] The 3D interactive visualization interface displays model node topology, data flow paths, and conflict warnings;

[0072] The verification system includes static verification and dynamic verification, and supports model version backtracking and evolution deduction.

[0073] Among them, the personalized features of the multi-heterogeneous sub-models to be integrated include the logical rules of the multi-heterogeneous sub-models, the constraints of the demand model, the topological structure of the architecture model, the parameterized features of the design model, and the manufacturing knowledge of the process model.

[0074] The following steps are involved in parsing the multi-heterogeneous sub-models to be fused, obtaining their personalized features, and constructing a unified semantic representation system based on the metamodel framework:

[0075] The model parsing engine uses differentiated parsing strategies for different model types. It extracts the functional chain and logic rules of SysML models and identifies concurrent execution logic in activity diagrams through semantic analysis. It also performs structured parsing of requirement models and uses natural language processing to extract requirement items and their constraints. It uses the Modelica interface to parse the topological connections and physical equations of the architecture model and extract state variables and parameter dependencies in differential algebraic equations. It uses the CAD software API to extract the geometric features, parametric constraints, and assembly relationships of the design model. Furthermore, it extracts manufacturing processes, equipment parameters, and process constraints from the process knowledge base to construct manufacturing feasibility assessment indicators for the process model.

[0076] A multi-dimensional meta-model framework is constructed based on the ontology language to uniformly represent the core attributes of heterogeneous models. The dimensions of the ontology language include functional dimension, structural dimension, behavioral dimension, domain semantics, and interface specification level.

[0077] The parsed model metadata is converted into standardized RDF triples and stored in the Neo4j graph database to form a unified data foundation that supports cross-model fusion and conversion.

[0078] Among them, according to the multivariate heterogeneous sub-models to be fused with standardized expressions, based on the knowledge graph and graph neural network, constructing a multivariate heterogeneous model based on the multivariate heterogeneous sub-models to be fused includes the following steps:

[0079] Differentiated parsing algorithms are used for different model types. The BERT+BiLSTM model is used to perform intent recognition and context analysis on the requirement text, extract the requirement constraint vector, and generate function labels.

[0080] Graph convolutional networks are used to learn the graph structure of SysML activity diagrams, capturing the timing dependencies and parallel execution characteristics in functional logic. Tensor decomposition technology is used to reduce the dimensionality of the topological connection relationships in the architecture model and encode them into a low-dimensional dense matrix to represent the structural characteristics.

[0081] Construct a five-dimensional feature matrix covering functions, structures, behaviors, domain semantics, and interface specifications, where the feature vector of each model instance contains functional labels, structural features, dynamic behavior attributes, domain semantic indicators, and interface compatibility parameters;

[0082] Cross-model association analysis is achieved through a semantic similarity calculation model. A weighted fusion strategy is adopted to integrate the cosine similarity of functional features, the Jaccard similarity of structural features, and the KL divergence of domain semantics to generate the semantic association strength value between models. The weight coefficients (α, β, γ) are dynamically optimized through a genetic algorithm to adapt to the model fusion conversion and coordination requirements in different application scenarios, ultimately forming a cross-model semantic association network that supports fusion conversion.

[0083] Among them, according to the multivariate heterogeneous sub-models to be fused with standardized expressions, based on the knowledge graph and the graph neural network, constructing a multivariate heterogeneous model based on the multivariate heterogeneous sub-models to be fused also includes the following steps:

[0084] Through the intermediate semantic layer, parameter mapping function library and interface adaptation engine, dynamic semantic alignment and interface interoperability between multi-source heterogeneous models are achieved. Timestamp synchronization and dynamic buffer scheduling are used to ensure the timing consistency of data interaction, realizing plug-and-play across model interfaces.

[0085] Among them, according to the multivariate heterogeneous sub-models to be fused with standardized expressions, based on the knowledge graph and the graph neural network, constructing a multivariate heterogeneous model based on the multivariate heterogeneous sub-models to be fused also includes the following steps:

[0086] With the system model as the root node, the requirement model as the constraint node, the architecture model as the structure node, the design model as the leaf node, and the process model as the end node, a hierarchical knowledge graph is constructed, and four types of associations are defined: requirement coverage edge, structure mapping edge, process constraint edge, and dynamic evolution edge.

[0087] The GraphSAGE algorithm is used to embed graphs and generate low-dimensional vectors to represent the functional coupling, structural dependency and semantic similarity between models.

[0088] To address the difficulty in collaborating multi-source heterogeneous models, including system models, requirement models, architecture models, design models, and process models, throughout the lifecycle of complex R&D systems, we build an automated semantic alignment mechanism based on a multi-level metamodel and an automated mapping mechanism between heterogeneous models based on a dynamic rule generation engine, thereby achieving semantic unification and dynamic interoperability between models. Specifically, while retaining the independent characteristics of each model, we use a hierarchical fusion mechanism to adaptively transform the parameter systems, interface specifications, and semantic logic of multi-source heterogeneous models. The specific process and implementation are as follows:

[0089] Multi-source heterogeneous model access and unified modeling: The model parsing engine extracts the logical rules of the system model, the constraints of the requirement model, the topological structure of the architecture model, the parameterized features of the design model, and the manufacturing knowledge of the process model, thereby building a unified semantic representation system based on the metamodel framework.

[0090] Multi-dimensional feature analysis and semantic mapping: Deep semantic parsing technology is used to extract features from the input and output dimensions, parameter constraints, and functional behavior characteristics of various models, and to establish a multi-dimensional feature matrix containing domain knowledge labels and dynamic evolution relationships;

[0091] Dynamic conversion layer construction: Develop a conversion architecture based on the intermediate semantic layer. Through parameter constraint mapping functions, interface specification conversion matrices, and semantic logic alignment modules, achieve target adaptation between system models and requirement models, parameter association between architecture models and design models, and dynamic matching between process models and manufacturing data.

[0092] Cross-model association network construction: Using graph embedding technology to mine the requirements coverage of system models, the design constraint chain of architecture models, and the manufacturing feasibility association of process models, we build a model collaborative relationship network covering the entire chain of "requirements-architecture-design-process, etc."

[0093] Full-link visualization and verification: Establish a multi-model fusion verification system, dynamically display the model interaction status, parameter transfer path and collaborative calculation confidence through a three-dimensional visualization interface, and support model version backtracking and evolution process deduction.

[0094] This method breaks through the data silos and semantic barriers in traditional model integration, supports dynamic semantic interoperability and adaptive collaboration of multi-source heterogeneous models throughout the life cycle of complex systems, and has significant technical advantages and application value in the field of intelligent equipment research and development.

[0095] Example of a fusion method for multi-source heterogeneous models

[0096] To address the difficulty in collaborating multi-source heterogeneous models, including system models, requirement models, architecture models, design models, and process models, throughout the lifecycle of complex R&D systems, we build an automated semantic alignment mechanism based on a multi-level metamodel and an automated mapping mechanism between heterogeneous models based on a dynamic rule generation engine, thereby achieving semantic unification and dynamic interoperability between models. Specifically, while retaining the independent characteristics of each model, we use a hierarchical fusion mechanism to adaptively transform the parameter systems, interface specifications, and semantic logic of multi-source heterogeneous models. The specific process and implementation are as follows:

[0097] (1) Multi-source heterogeneous model access and unified modeling

[0098] In this phase, the model parsing engine and unified meta-model framework are used to achieve standardized access and semantic unification of system models (SysML / UML), requirement models, architecture models (Modelica), design models (CAD parametric models), and process models (process BOM / manufacturing knowledge base). First, the model parsing engine adopts differentiated parsing strategies for different model types, extracting the functional chain and logical rules of the SysML model and identifying the concurrent execution logic in the activity diagram through semantic analysis; performing structured parsing of the requirement model and extracting the requirement items and their constraints using natural language processing technology; parsing the topological connection relationship and physical equations of the architecture model through the Modelica interface, extracting the state variables and parameter dependencies in the differential algebraic equations; extracting the geometric features, parametric constraints, and assembly relationships of the design model based on the CAD software's API interface; and extracting manufacturing procedures, equipment parameters, and process constraints from the process knowledge base to construct manufacturing feasibility evaluation indicators for the process model. On this basis, a five-dimensional metamodel framework was constructed based on ontology language. This framework uniformly represents the core attributes of heterogeneous models from five perspectives: functional dimension (logical rules and target constraints), structural dimension (topological connections and parameter chains), behavioral dimension (dynamic evolution equations), domain semantics (knowledge labels), and interface specifications (data types and communication protocols). Ultimately, the parsed model metadata was converted into standardized RDF triples and stored in the Neo4j graph database, forming a unified data foundation that supports cross-model fusion and transformation.

[0099] (2) Multidimensional feature analysis and semantic mapping

[0100] This phase achieves cross-model semantic alignment and relationship mining through deep semantic parsing technology and multi-dimensional feature matrix construction. Differentiated parsing algorithms are employed for different model types. The BERT+BiLSTM model is used to perform intent recognition and context analysis on the requirement text, extracting the requirement constraint vector and generating functional labels. A graph convolutional network (GCN) is used to learn the graph structure of SysML activity diagrams, capturing the temporal dependencies and parallel execution characteristics of functional logic. Tensor decomposition technology is combined to reduce the dimensionality of the topological connection relationships of the architectural model and encode them into a low-dimensional dense matrix to represent the structural features. On this basis, a five-dimensional feature matrix covering function, structure, behavior, domain semantics, and interface specifications is constructed. The feature vector of each model instance includes functional labels (e.g., requirement coverage, logical completeness score), structural features (e.g., node connectivity, parameter coupling), dynamic behavioral attributes (e.g., state variable dimensions), domain semantic indicators (e.g., manufacturing feasibility score, process constraint conflict value), and interface compatibility parameters (e.g., data type matching). Cross-model association analysis is further implemented through a semantic similarity calculation model. A weighted fusion strategy integrates cosine similarity of functional features, Jaccard similarity of structural features, and KL divergence of domain semantics to generate semantic association strength values ​​between models. The weight coefficients (α, β, γ) are dynamically optimized using a genetic algorithm to adapt to the collaborative needs of model fusion and transformation in different application scenarios, ultimately forming a cross-model semantic association network that supports fusion and transformation.

[0101] (3) Dynamic conversion layer construction

[0102] This phase achieves dynamic semantic alignment and interface interoperability between multi-source heterogeneous models through an intermediate semantic layer (ISL), a parameter mapping library, and an interface adaptation engine. The intermediate semantic layer defines a unified intermediate representation language (IRL), converting the logical expressions of different models into a standardized form of IRL. For example, SysML's "concurrent execution" logic is mapped into an IRL Petri net model, which is then converted into a Modelica model using semantic bridging rules. The parameter mapping library establishes dynamic conversion rules for scalar parameters (such as temperature unit conversion), multidimensional parameters (dimensionality reduction matching based on principal component analysis), and constraint parameters (embedding requirement inequalities into design goals), resolving the issue of mismatches between numerical types and dimensions. The interface adaptation engine, developed based on the ROS middleware, supports real-time conversion between multiple protocols and formats (JSON, XML). Timestamp synchronization and dynamic buffer scheduling ensure temporal consistency of data interactions, enabling plug-and-play cross-model interfaces.

[0103] (4) Construction of cross-model association network

[0104] In this phase, a model collaboration relationship network covering the entire life cycle is constructed through knowledge graph and graph neural network technology. With the system model as the root node, the demand model as the constraint node, the architecture model as the structural node, the design model as the leaf node, and the process model as the terminal node, a hierarchical knowledge graph is constructed, and four types of association relationships are defined: demand coverage edge (weight = demand satisfaction), structure mapping edge (weight = parameter matching), process constraint edge (weight = manufacturing feasibility), and dynamic evolution edge (version iteration relationship). On this basis, the GraphSAGE algorithm is used to embed the graph and generate low-dimensional vectors to represent the functional coupling, structural dependency, and semantic similarity between models, providing a relational reasoning basis for intelligent fusion and supporting the automated recommendation of model collaboration strategies.

[0105] (5) Full-link visualization and verification

[0106] In this phase, the reliability and explainability of model fusion are ensured through a visualization engine and a multi-dimensional verification system. An interactive visualization interface is developed based on Unity3D to dynamically display the model node topology (color coding to distinguish model types), data flow paths (highlighting key parameter transfer links), and conflict warnings (flashing red to indicate interface or semantic anomalies). The verification system includes static verification (using the OWL reasoning engine to detect logical consistency) and dynamic verification, supporting model version backtracking and evolutionary deduction, ensuring that the entire collaborative process from demand to manufacturing is traceable and verifiable.

[0107] (1) Dynamic conversion and interface adaptation technology for multi-source heterogeneous models: The present invention proposes a method and system for dynamic conversion and interface adaptation of multi-source heterogeneous models, aiming to solve the problems of logical expression differences, parameter mismatches and poor interface interoperability between different models. By defining a unified intermediate representation language (IRL), the logical expressions of different models such as SysML's "concurrent execution" logic are converted into a standardized form of IRL, such as a Petri net model mapped to IRL, and then converted into a Modelica model through semantic bridging rules. At the same time, a parameter mapping function library is established to formulate dynamic conversion rules for scalar parameters, multi-dimensional parameters and constraint parameters to solve the problem of mismatch between numerical types and dimensions. In addition, the interface adaptation engine developed based on ROS middleware supports real-time conversion of multiple protocols and multiple formats, and ensures the timing consistency of data interaction through timestamp synchronization and dynamic buffer scheduling, realizing plug-and-play of cross-model interfaces.

[0108] (2) Full-link visualized model fusion verification technology: The present invention also relates to a full-link visualized model fusion verification method and system, which is dedicated to improving the reliability and interpretability of model fusion. The interactive visualization interface developed based on Unity3D can dynamically display the model node topology (color coding to distinguish model types), data flow path (highlight key parameter transfer links) and conflict warning (flashing red to prompt interface or semantic anomalies). At the same time, a multi-dimensional verification system including static verification (detecting logical consistency through OWL reasoning engine) and dynamic verification is constructed, which supports model version backtracking and evolutionary deduction, ensuring that the full-link collaborative process from demand to manufacturing is traceable and verifiable.

[0109] The highlight of the present invention lies in its efficient collaboration and dynamic adaptation capabilities, as well as its innovative mechanism for dynamic conversion across models. Specifically, by constructing an automated semantic alignment mechanism based on a multi-level metamodel, the present invention achieves efficient collaboration of multi-source heterogeneous models such as system models, demand models, architecture models, design models, and process models. This mechanism can effectively solve the problems of low efficiency in collaboration between models and poor dynamic adaptation capabilities in traditional integration methods. At the same time, the use of a dynamic rule generation engine can cope with dynamic scenarios such as demand changes and parameter iterations, solving the limitations of traditional integration methods that rely on static mapping rules.

[0110] In terms of dynamic conversion across models, the present invention realizes dynamic semantic alignment and interface interoperability between multi-source heterogeneous models through an intermediate semantic layer (ISL), a parameter mapping function library, and an interface adaptation engine. The intermediate semantic layer defines a unified intermediate representation language (IRL) to convert the logical expressions of different models into a standardized form of IRL. The interface adaptation engine is developed based on ROS middleware, supports real-time conversion of multiple protocols and multiple formats (JSON, XML), and ensures the timing consistency of data interaction through timestamp synchronization and dynamic scheduling of buffers. This design not only realizes plug-and-play of cross-model interfaces, but also ensures the efficiency and reliability of data interaction, greatly improving the overall performance of the system.

[0111] Fusion method and device for multi-source heterogeneous models

[0112] See attached Figure 2 The fusion device of the multivariate heterogeneous model provided by the embodiment of the present invention includes:

[0113] Multi-source heterogeneous sub-model acquisition module, used to obtain multi-source heterogeneous sub-models to be fused;

[0114] A unified semantic representation system construction module is used to parse the multi-heterogeneous sub-models to be fused, obtain the personalized features of the multi-heterogeneous sub-models to be fused, and construct a unified semantic representation system based on the metamodel framework;

[0115] A standardized expression module is used to standardize the logical expressions of the multivariate heterogeneous sub-models to be fused using a unified intermediate representation language according to the personalized characteristics of the multivariate heterogeneous sub-models to be fused, thereby obtaining the multivariate heterogeneous sub-models to be fused in a standardized expression form;

[0116] The multivariate heterogeneous model construction module is used to construct a multivariate heterogeneous model based on the multivariate heterogeneous sub-models to be fused based on the knowledge graph and graph neural network according to the multivariate heterogeneous sub-models to be fused with standardized expressions.

[0117] Computer-readable storage medium

[0118] The computer-readable storage medium provided by the present invention stores a fusion program of multi-source heterogeneous models. When the fusion program of multi-source heterogeneous models is executed by a processor, the steps of the fusion method of multi-source heterogeneous models provided by the present invention are implemented.

[0119] electronic devices

[0120] The electronic device provided by the present invention includes a memory and a storage, and a fusion program of multi-source heterogeneous models is stored in the memory. When the fusion program of multi-source heterogeneous models is executed by the processor, the steps of the fusion method of multi-source heterogeneous models provided by the present invention are implemented.

[0121] Reference Figure 3 , Figure 3 This is a schematic diagram of the device structure of a method for fusing multi-source heterogeneous models of a hardware operating environment involved in an embodiment of the present invention.

[0122] like Figure 3 As shown, the multi-source heterogeneous model fusion method and device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to implement connection and communication between these components. The user interface 1003 may include a display screen (Display) and an input unit such as a keyboard (Keyboard). Optionally, the user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a wireless fidelity (WI-FI) interface). The memory 1005 may be a high-speed random access memory (RAM) memory or a stable non-volatile memory (NVM), such as a disk memory. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0123] Those skilled in the art will understand that Figure 3 The structure shown in does not constitute a limitation on the multi-source heterogeneous model fusion method and equipment, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0124] like Figure 3 As shown, the memory 1005 as a storage medium may include an operating system, a data storage module, a network communication module, a user interface module, and a multi-source heterogeneous model fusion method program.

[0125] exist Figure 3 In the multi-source heterogeneous model fusion method device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the multi-source heterogeneous model fusion method device of the present invention can be set in the multi-source heterogeneous model fusion method device, and the multi-source heterogeneous model fusion method device calls the multi-source heterogeneous model fusion method program stored in the memory 1005 through the processor 1001, and executes the multi-source heterogeneous model fusion method provided by the embodiment of the present invention.

[0126] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0127] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A method for fusing multi-source heterogeneous models, characterized in that: The following steps are involved: Obtaining the multivariate heterogeneous sub-models to be fused; Analyze the multi-heterogeneous sub-models to be fused, obtain personalized features of the multi-heterogeneous sub-models to be fused, and construct a unified semantic representation system based on the metamodel framework; According to the personalized characteristics of the multi-heterogeneous sub-models to be fused, the logical expressions of the multi-heterogeneous sub-models to be fused are standardized using a unified intermediate representation language to obtain the multi-heterogeneous sub-models to be fused with a standardized expression form; According to the multivariate heterogeneous sub-models to be fused with standardized expressions, a multivariate heterogeneous model based on the multivariate heterogeneous sub-models to be fused is constructed based on the knowledge graph and graph neural network.

2. The fusion method of multi-source heterogeneous models according to claim 1, characterized in that: The following steps are also included: Based on the three-dimensional interactive visualization interface, a visualization engine and multi-dimensional verification system are constructed, including: The 3D interactive visualization interface displays model node topology, data flow paths, and conflict warnings; The verification system includes static verification and dynamic verification, and supports model version backtracking and evolution deduction.

3. The fusion method of multi-source heterogeneous models according to claim 1, characterized in that: The personalized features of the multi-heterogeneous sub-models to be integrated include the logical rules of the multi-heterogeneous sub-models, the constraints of the demand model, the topological structure of the architecture model, the parameterized features of the design model, and the manufacturing knowledge of the process model.

4. The fusion method of multi-source heterogeneous models according to claim 1, characterized in that: Parsing the multi-heterogeneous sub-models to be fused to obtain personalized features of the multi-heterogeneous sub-models to be fused, and constructing a unified semantic representation system based on the metamodel framework specifically includes the following steps: The model parsing engine uses differentiated parsing strategies for different model types. It extracts the functional chain and logic rules of SysML models and identifies concurrent execution logic in activity diagrams through semantic analysis. It also performs structured parsing of requirement models and uses natural language processing to extract requirement items and their constraints. It uses the Modelica interface to parse the topological connections and physical equations of the architecture model and extract state variables and parameter dependencies in differential algebraic equations. It uses the CAD software API to extract the geometric features, parametric constraints, and assembly relationships of the design model. Furthermore, it extracts manufacturing processes, equipment parameters, and process constraints from the process knowledge base to construct manufacturing feasibility assessment indicators for the process model. Constructing a multidimensional meta-model framework based on ontology language to uniformly represent the core attributes of heterogeneous models, wherein the dimensions of the ontology language include functional dimension, structural dimension, behavioral dimension, domain semantics, and interface specification level; The parsed model metadata is converted into standardized RDF triples and stored in the Neo4j graph database to form a unified data foundation that supports cross-model fusion and conversion.

5. The fusion method of multi-source heterogeneous models according to claim 1, characterized in that: According to the multivariate heterogeneous sub-models to be fused in a standardized expression form, constructing a multivariate heterogeneous model based on the multivariate heterogeneous sub-models to be fused based on a knowledge graph and a graph neural network includes the following steps: Differentiated parsing algorithms are used for different model types. The BERT+BiLSTM model is used to perform intent recognition and context analysis on the requirement text, extract the requirement constraint vector, and generate function labels. Graph convolutional networks are used to learn the graph structure of SysML activity diagrams, capturing the timing dependencies and parallel execution characteristics in functional logic. Tensor decomposition technology is used to reduce the dimensionality of the topological connection relationships in the architecture model and encode them into a low-dimensional dense matrix to represent the structural characteristics. Construct a five-dimensional feature matrix covering functions, structures, behaviors, domain semantics, and interface specifications, where the feature vector of each model instance contains functional labels, structural features, dynamic behavior attributes, domain semantic indicators, and interface compatibility parameters; Cross-model association analysis is achieved through a semantic similarity calculation model. A weighted fusion strategy is adopted to integrate the cosine similarity of functional features, the Jaccard similarity of structural features, and the KL divergence of domain semantics to generate the semantic association strength value between models. The weight coefficients (α, β, γ) are dynamically optimized through a genetic algorithm to adapt to the model fusion conversion and coordination requirements in different application scenarios, ultimately forming a cross-model semantic association network that supports fusion conversion.

6. The method for fusing multi-source heterogeneous models according to claim 5, characterized in that: According to the multivariate heterogeneous sub-models to be fused in a standardized expression form, constructing a multivariate heterogeneous model based on the multivariate heterogeneous sub-models to be fused based on a knowledge graph and a graph neural network further includes the following steps: Through the intermediate semantic layer, parameter mapping function library and interface adaptation engine, dynamic semantic alignment and interface interoperability between multi-source heterogeneous models are achieved. Timestamp synchronization and dynamic buffer scheduling are used to ensure the timing consistency of data interaction, realizing plug-and-play across model interfaces.

7. The method for fusing multi-source heterogeneous models according to claim 5, characterized in that: According to the multivariate heterogeneous sub-models to be fused in a standardized expression form, constructing a multivariate heterogeneous model based on the multivariate heterogeneous sub-models to be fused based on a knowledge graph and a graph neural network further includes the following steps: With the system model as the root node, the requirement model as the constraint node, the architecture model as the structure node, the design model as the leaf node, and the process model as the end node, a hierarchical knowledge graph is constructed, and four types of associations are defined: requirement coverage edge, structure mapping edge, process constraint edge, and dynamic evolution edge. The GraphSAGE algorithm is used to embed graphs and generate low-dimensional vectors to represent the functional coupling, structural dependency and semantic similarity between models.

8. A fusion device for multi-element heterogeneous models, characterized in that: include: Multi-source heterogeneous sub-model acquisition module, used to obtain multi-source heterogeneous sub-models to be fused; A unified semantic representation system construction module is used to parse the multi-heterogeneous sub-models to be fused, obtain personalized features of the multi-heterogeneous sub-models to be fused, and construct a unified semantic representation system based on the metamodel framework; A standardized expression module is used to standardize the logical expressions of the multivariate heterogeneous sub-models to be fused according to the personalized characteristics of the multivariate heterogeneous sub-models to be fused using a unified intermediate representation language to obtain the multivariate heterogeneous sub-models to be fused with a standardized expression form; A multivariate heterogeneous model construction module is used to construct a multivariate heterogeneous model based on the multivariate heterogeneous sub-models to be fused based on the knowledge graph and graph neural network according to the multivariate heterogeneous sub-models to be fused with standardized expressions.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a fusion program of multiple-source heterogeneous models. When the fusion program of multiple-source heterogeneous models is executed by the processor, the steps of the fusion method of multiple-source heterogeneous models described in any one of claims 1 to 7 are implemented.

10. An electronic device, characterized in that: It includes a memory and a storage, wherein a fusion program of multi-source heterogeneous models is stored on the memory, and when the fusion program of multi-source heterogeneous models is executed by the processor, the steps of the fusion method of multi-source heterogeneous models described in any one of claims 1-7 are implemented.

Images