A multi-view model automatic generation method adapting to an MBSE architecture standard

Abstract

This invention discloses an automatic multi-view model generation method adapted to the MBSE architecture standard, belonging to the technical field of automatic multi-view model generation. This invention constructs a four-level scalable view rule system, uses a bidirectional matching algorithm to calculate forward and reverse matching scores, and weights them to obtain the total matching degree. It automatically determines the target view set matching the target long text document, then performs batch multi-view modeling through multi-threaded parallel processing. For each individual view data, it calculates the element-level deviation, view-level compliance score, and view topology association weight coefficient to obtain a consistency evaluation index. After correcting conflicts, it outputs a standardized model file, achieving end-to-end automation from design scheme text input to standardized model file output. This solves the problems of low efficiency, error-proneness, and poor standard compatibility caused by relying on manual specification of view types and manual modeling in existing technologies.

Description

Technical Field

[0001] This invention relates to the field of automatic multi-view model generation technology, and in particular to a method for automatic multi-view model generation adapted to the MBSE architecture standard. Background Technology

[0002] Model-based systems engineering (MBSE) in the development of complex equipment requires the generation of multiple view models following architecture standards such as DoDAF and UAF. Currently, the process of generating these standard view models heavily relies on professionals familiar with specific architecture standards. They manually read design documents, identify modeling elements, business constraints, and design requirements, and then build each view one by one using modeling tools such as Cameo Systems Modeler. This manual approach not only results in lengthy system modeling cycles for single-model equipment, making it difficult to adapt to the rapid iteration needs of equipment development, but also allows for discrepancies in understanding of standards and specifications among different personnel, potentially leading to data inconsistencies between models and creating logical conflicts in multi-system collaboration and top-level system planning scenarios.

[0003] In related technologies, there are some semi-automated assisted modeling tools for single view types. These tools require manual pre-specification of the specific view type to be generated, then extract simple features from the design document based on a fixed template, and finally map them onto the view elements. For example, users need to know in advance whether to generate a CV-1 view or an OV-1 view under the DoDAF standard and manually select the corresponding generation template. In addition, existing methods are usually deeply bound to specific system architecture standard versions, and the generated model file formats are not uniform, making them difficult to be compatible with modeling tools from different vendors.

[0004] However, the aforementioned existing technologies cannot achieve end-to-end automated generation from equipment design documents to a complete set of compliant view models. Specifically, this manifests in the following ways: First, the type of view to be generated must be manually specified in advance, as the system lacks autonomous judgment capabilities; second, there is a lack of intelligent matching mechanisms between text parsing and view generation, resulting in the inability to automatically transmit business constraints and performance indicators in the design scheme to the corresponding view models; third, cross-view consistency assurance relies entirely on manual verification, lacking automated conflict detection and correction methods. Summary of the Invention

[0005] This application provides an automatic multi-view model generation method that adapts to the MBSE architecture standard. This solves the problems of low efficiency, error-proneness and poor standard compatibility in the prior art, which relies on manually specifying view types and manually modeling. It achieves end-to-end automation from design scheme text input to standardized model file output.

[0006] This application provides a method for automatically generating multi-view models that adapts to the MBSE architecture standard, including:

[0007] S1. Obtain the view rule system that adapts to the MBSE architecture standard. The view rule system is a set of view rules that includes view type characteristics, view generation trigger conditions, view standard specification requirements, and business topology dependencies between views.

[0008] S2 performs structured parsing on the input target long text document, extracting modeling elements, business constraints, and design requirements from the target long text document to obtain standardized modeling feature data;

[0009] S3, by matching and analyzing the standardized modeling feature data with the view rule system, the types and number of views to be generated that match the target long text document are automatically determined, and the view generation sequence is automatically planned in combination with the business topology dependency relationship between views to obtain the target view set;

[0010] S4. By calling the view standard specifications in the view rule system, batch multi-view modeling is automatically performed according to the target view set, and conflicts are corrected through cross-view element consistency verification to obtain the multi-view model.

[0011] S5. Perform standardized format conversion on the multi-view model and output a standardized model file that conforms to the MBSE modeling tool compatible format.

[0012] Furthermore, the view rule system adopts a four-level extensible architecture, including a standard version layer, a view group layer, a view layer, and a rule item layer;

[0013] The standard version layer is the adapted MBSE architecture version number, used to adapt to different versions of MBSE standard specifications;

[0014] The view group layer is a collection of views divided by business domain, used to classify and manage views within the same business domain;

[0015] The view layer is a specific MBSE single-view data type, corresponding to the smallest unit of single-view data modeling;

[0016] Each rule item in the rule item layer is defined using a four-tuple structure. The four-tuple includes the view generation trigger condition, feature keywords, view standard specification requirements, and business topology dependencies between views, which are used to define the view generation rules.

[0017] Furthermore, the standardized modeling feature data is matched and analyzed with the view rule system, specifically including the following steps:

[0018] S31, Map the standardized modeling feature data to the view rule system using feature vectorization to obtain feature vectors and rule vectors, calculate the matching degree between feature vectors and rule vectors, and filter the views whose matching degree reaches the preset threshold to obtain a set of candidate views;

[0019] S32, perform constraint conflict verification and MBSE standard compliance verification on the candidate view set, remove conflicting and non-compliant views, obtain the types and number of views to be generated that match the target long text document, and automatically plan the view generation sequence based on the business topology dependencies between views to obtain the target view set.

[0020] Furthermore, the matching degree is calculated using a bidirectional matching algorithm, the specific steps of which include:

[0021] S311, Calculate the positive matching score between the feature vector and the rule vector. The positive matching score is used to characterize the overall similarity between the feature vector and the rule vector. The formula for calculating the positive matching score is:

[0022] ;

[0023] in, For positive matching scores, This refers to the number of dimensions in the feature vector, i.e., the number of preset fields for the modeling features. The summation index variable represents the i-th dimension of the feature vector. The normalized quantized value of the feature vector in the i-th dimension. Let be the normalized weight value of the rule vector in the i-th dimension;

[0024] S312, calculate the back-matching score between the mandatory constraint vector in the rule term and the associated dimension subset of the corresponding mandatory constraint in the feature vector. The back-matching score is used to characterize the completeness of the mandatory constraint requirements in the rule term being covered by the standardized modeling feature data. The formula for calculating the back-matching score is:

[0025] ;

[0026] in, For reverse matching scores, This represents the total number of dimensions for mandatory constraints in the current view's rule items, where j is the dimension index of the mandatory constraint item. Let J be the normalized weight of the j-th dimension mandatory constraint term in the rule term. This is the normalized quantized value of the dimension associated with the j-th dimension of the mandatory constraint term in the feature vector;

[0027] S313, the forward matching score and the backward matching score are weighted and summed to obtain the total matching degree. ,in, The weights for positive matching scores, The weights for the reverse matching score, ;

[0028] S314, filter the views whose total matching degree reaches the preset threshold to obtain a set of candidate views.

[0029] Furthermore, the constraint conflict verification and MBSE standard compliance verification of the candidate view set includes the following steps:

[0030] S321, Extract business constraints and performance constraints from the target long text document;

[0031] S322, Constraint Conflict Validation: The extracted business constraints and performance constraints are compared with the rule requirements of the corresponding candidate views in the view rule system. Combined with the predefined business topology dependencies between views in the view rule system, conflicting views with logical mutual exclusion or missing dependencies in the candidate view set are determined and eliminated.

[0032] S323, MBSE Standard Compliance Verification: Based on the mandatory specifications of the MBSE architecture standard, remove views with standard violations from the candidate view set;

[0033] S324. After both constraint conflict verification and standard compliance verification are completed, sort the remaining views in the candidate view set from high to low matching degree, and retain the top N views with the highest matching degree.

[0034] S325, combining the business topology dependencies between views, automatically plans the view generation sequence to obtain the target view set.

[0035] Furthermore, the specific steps of the batch multi-view modeling include:

[0036] S411, decompose the target view set in the task dimension to obtain distributed modeling subtasks corresponding to multiple single view generation tasks;

[0037] S412, the distributed modeling subtask is executed through multi-threaded parallel processing, with each thread completing a modeling task for a single view data item and generating the single view data items;

[0038] S413 performs a three-level cross-view feature consistency check on each individual view data at the feature level, view level, and system level. If the consistency check passes, the individual view data is directly output as the corrected view feature.

[0039] S414 When the data of a single view fails the consistency check, a baseline view is determined according to the preset priority rules, and the attribute values ​​of the corresponding inconsistent elements in other views are uniformly corrected to the attribute values ​​of the element in the baseline view to obtain the corrected view features.

[0040] S415 maps the corrected view features to preset view visualization parameters element by element to complete the visualization rendering and obtain a multi-view model.

[0041] Furthermore, the cross-view feature consistency verification of each individual view data includes: checking whether the attribute values ​​of the same modeling feature are consistent in different views, and checking whether each individual view data conforms to the metamodel specification of the MBSE architecture standard. Specific steps include:

[0042] S4131, For the numerical modeling features of each individual view data in the multi-view model, calculate the feature-level deviation D. The formula for calculating the feature-level deviation D is: ,in , Attribute values ​​of the same numerical modeling feature in different views. The maximum permissible attribute deviation value preset for this modeling element in the MBSE system architecture standard is used for deviation determination in consistency verification. The value range of D is [0,1].

[0043] S4132, For the number of rule items in each individual view data in the multi-view model, calculate the view-level compliance score C. The formula for calculating the view-level compliance score C is as follows: ,in This represents the number of rule items in the single-view data that conform to the MBSE standard specification. The total number of rule items in this single view data, C, has a value range of [0,1].

[0044] S4133, Based on the business topology dependency relationship between views, obtain the preset view topology association weight coefficient W for each individual view data, and the sum of the view topology association weight coefficients of all views is 1.

[0045] S4134, For each individual view data item, calculate the consistency evaluation index Q. ,in, , , These are the weighting coefficients corresponding to feature-level deviation, view-level compliance score, and preset view topology association weighting coefficients, respectively. , , All are positive numbers, and ;

[0046] S4135, Obtain the preset convergence threshold. When consistency evaluation indicators When the data in a single view passes the consistency check, it is determined that the data in that single view passes the consistency check. If the data in a single view fails the consistency check, step S414 is triggered.

[0047] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0048] This application provides an automatic multi-view model generation method adapted to the MBSE architecture standard. The method obtains a view rule system adapted to the MBSE architecture standard, which adopts a four-level scalable architecture. Rule items define view generation trigger conditions, feature keywords, view standard specification requirements, and business topology dependencies between views using quadruples. The method performs structured parsing of the target long text document, extracting modeling elements, business constraints, and design requirements to obtain standardized modeling feature data. Feature vectors and rule vectors are obtained through feature vectorization mapping. Forward and reverse matching scores are calculated and weighted to obtain the total matching degree. Views reaching a preset threshold are selected to obtain a candidate view set. Constraint conflict verification and MBSE standard compliance verification are performed, eliminating conflicting and non-compliant views. A view generation sequence is planned based on the business topology dependencies between views to obtain the target view set. Batch multi-view modeling is performed, generating individual view data through multi-threaded parallel processing. Cross-view element consistency verification is performed, calculating element-level deviation, view-level compliance score, and view topology association weight coefficient to obtain consistency evaluation indicators. After conflict correction, visualization rendering is completed to obtain the multi-view model. Finally, standardized format conversion is performed to output a standardized model file.

[0049] In this process, a four-level scalable view rule system is used to structurally store the view generation trigger conditions, feature keywords, standard specification requirements, and business topology dependencies between views under different versions of the MBSE standard, enabling matching analysis to be performed based on a unified rule description. Through structured parsing of the target long text document, modeling elements, business constraints, and design requirements are automatically extracted, eliminating the need for manual reading and input, thus lowering the professional threshold for modeling.

[0050] Furthermore, feature vectors and rule vectors are obtained through feature vectorization mapping. Forward and reverse matching scores are calculated, and a weighted sum is obtained to obtain the total matching degree. Then, constraint conflict checks and MBSE standard compliance checks are performed to eliminate views with logical mutual exclusion and standard violations. Finally, the view generation sequence is planned based on the business topology dependencies between views. This process eliminates the need for manual specification of view types, achieving automatic matching from design text to the target view set, and avoiding view omissions or redundancy caused by manual judgment.

[0051] Furthermore, in batch multi-view modeling, individual view data is generated through multi-threaded parallel processing. For each individual view, feature-level deviation, view-level compliance score, and view topology association weight coefficient are calculated to obtain a consistency evaluation index. When the consistency evaluation index reaches the convergence threshold, the model is considered successful; otherwise, a correction operation is triggered, adjusting the attribute values ​​of the feature in other views to match the attribute values ​​in the baseline view. This quantitative cross-view feature consistency verification automatically detects and corrects attribute deviations of the same modeled feature in different views, as well as deviations of individual view data from standard specifications, ensuring the reliability of the multi-view model. The final standardized model file can be directly read by mainstream MBSE modeling tools without manual format conversion. Attached Figure Description

[0052] Figure 1 A flowchart illustrating an automatic multi-view model generation method adapted to the MBSE architecture standard, provided in this application embodiment. Detailed Implementation

[0053] This application provides an automatic multi-view model generation method adapted to the MBSE architecture standard. It solves the problems of low efficiency, error-proneness and poor standard compatibility caused by relying on manual specification of view types and manual modeling in the prior art. By constructing a four-level scalable view rule system, a bidirectional matching algorithm is used to calculate the forward matching score and the reverse matching score and then weighted and summed to obtain the total matching degree. The target view set matching the target long text document is automatically determined. Then, batch multi-view modeling is performed through multi-threaded parallel processing. The element-level deviation degree, view-level compliance score and view topology association weight coefficient are calculated for each individual view data to obtain the consistency evaluation index. After correcting conflicts, a standardized model file is output, realizing end-to-end automation from design scheme text input to standardized model file output.

[0054] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0055] like Figure 1 As shown in the figure, this application provides a method for automatically generating multi-view models that adapts to the MBSE architecture standard, including:

[0056] S1. Obtain the view rule system that adapts to the MBSE architecture standard. The view rule system is a set of view rules that includes view type characteristics, view generation trigger conditions, view standard specification requirements, and business topology dependencies between views.

[0057] S2 performs structured parsing on the input target long text document, extracting modeling elements, business constraints, and design requirements from the target long text document to obtain standardized modeling feature data;

[0058] It should be explained that the target long text document refers to various text materials carrying information related to system modeling, including one or more of the following: equipment requirements documents, design scheme documents, and technical specification documents; for example, equipment design scheme documents, requirements specification documents, system construction schemes, top-level planning documents, technical specification documents, or operational scenario documents. Supported text formats for the target long text document include TXT, DOCX, PDF, and Markdown formats, as well as structured design data imported from third-party systems. Input methods for the target long text document include, but are not limited to: direct input into the system's interactive interface input box, local file upload and import, synchronization via third-party business system interfaces, and document retrieval from the cloud.

[0059] The structured parsing is performed using a multi-modal fusion strategy that combines a large language model with a domain knowledge graph. The large language model is used for semantic understanding and entity extraction of the text, and the domain knowledge graph is used to provide entity types and relation definitions in the field of equipment modeling.

[0060] During the parsing process, a structured modeling feature vector and a set of entity elements are output. Each dimension of the structured modeling feature vector corresponds to a normalized quantized value of a standardized modeling element. The set of entity elements includes at least one of organizational nodes, combat missions, system components, data links, and capability indicators.

[0061] The standardized modeling feature data includes: the structured modeling feature vector, the entity element set, and business constraints extracted from the text.

[0062] Unlike traditional single-view generation that extracts local features from a single view, this step addresses the need for batch generation of full views. It employs a multi-mode fusion parsing strategy to extract global features that can support full-domain modeling. Parsing methods include, but are not limited to, semantic parsing based on large language models, BERT vector feature vectorization extraction, rule matching, machine learning classification model classification, knowledge graph triple extraction, and OCR image-to-text parsing. The final output is a structured modeling feature vector and a set of entity elements, providing a unified feature foundation for subsequent view matching and batch modeling.

[0063] S3, by matching and analyzing the standardized modeling feature data with the view rule system, the types and number of views to be generated that match the target long text document are automatically determined, and the view generation sequence is automatically planned in combination with the business topology dependency relationship between views to obtain the target view set;

[0064] S4. By calling the view standard specifications in the view rule system, batch multi-view modeling is automatically performed according to the target view set, and conflicts are corrected through cross-view element consistency verification to obtain the multi-view model.

[0065] S5. Perform standardized format conversion on the multi-view model and output a standardized model file that conforms to the MBSE modeling tool compatible format.

[0066] It should be explained that after generating the multi-view model, the multi-view model is converted to a standardized format and output as a standardized model file in XMI format, which is compatible with mainstream MBSE modeling tools such as Rhapsody, MagicDraw, and Chufeng Modeling Platform.

[0067] In this embodiment, by constructing a view rule system and performing feature parsing on design scheme text, the target view set is intelligently determined and automated batch modeling is executed. This improves the automation level, standard adaptability, and tool ecosystem compatibility of MBSE system modeling, and is a key step supporting generative modeling from single-view, single-step generation to one-click full-view batch generation. This invention achieves fully automated MBSE modeling from equipment design scheme input to full-view output, eliminating the need for manual specification of view types, manual input of modeling elements, and DoDAF / UAF / UPDM professional qualifications. Even R&D personnel with no prior experience can quickly complete complex equipment system modeling. Modeling efficiency is improved compared to traditional manual methods, reducing manpower and time costs and adapting to the rapid iterative R&D needs of equipment. Furthermore, the multi-view model generated by this invention can be output as a standardized model file, which can be directly read and edited by mainstream MBSE modeling tools without manual format conversion. It can be effectively integrated into existing equipment digital R&D processes and has high engineering application and promotion value.

[0068] Furthermore, the view rule system adopts a four-level extensible architecture, including a standard version layer, a view group layer, a view layer, and a rule item layer;

[0069] The standard version layer is the adapted MBSE architecture version number, used to adapt to different versions of MBSE standard specifications;

[0070] The view group layer is a collection of views divided by business domain, used to classify and manage views within the same business domain;

[0071] The view layer is a specific MBSE single-view data type, corresponding to the smallest unit of single-view data modeling;

[0072] Each rule item in the rule item layer is defined using a four-tuple structure. The four-tuple includes the view generation trigger condition, feature keywords, view standard specification requirements, and business topology dependencies between views, which are used to define the view generation rules.

[0073] It should be explained that in the four-level scalable architecture, the view type characteristics are stored in the view layer, the view generation triggering conditions, view standard specification requirements, and business topology dependencies between views are stored in the rule item layer, and the standard version layer stores the architecture version number.

[0074] The four-tier scalable architecture, from top to bottom, is as follows:

[0075] The first level is the standard version layer of the architecture, which is natively compatible with all versions of DoDAF, UAF, and UPDM, and can be adapted to different versions of the MBSE architecture standard specifications.

[0076] The second level is the view group layer, which is a collection of views divided by business domain. It is used to classify and manage views within the same business domain, covering the full view group under the corresponding MBSE standard.

[0077] The third level is the view layer, which is the specific MBSE single-view data type. It corresponds to the smallest unit of single-view data modeling and covers all views under the corresponding standard.

[0078] Level 4 is the rule item layer, which includes four main categories of rule items: view generation trigger conditions, feature keywords, view standard specification requirements, and business topology dependencies between views.

[0079] The rule items are defined using a standardized four-tuple structure {Condition, Feature, Norm, Constraint}; the meaning and implementation of each tuple are as follows:

[0080] Where Condition is the trigger condition for view generation, which is implemented using semantic similarity thresholds or regular expressions; Feature is a predefined feature keyword, specifically referring to the SysML / DoDAF / UAF system feature keyword library pre-built according to the MBSE architecture standard; Norm corresponds to the meta-model specification defined in the OMG SysML and DoDAF / UAF architecture standards; Constraint uses logical expressions or SPARQL statements to define the business dependencies between views.

[0081] The view generation triggering conditions are implemented using semantic similarity thresholds or regular expressions.

[0082] The predefined feature keyword library is a vocabulary set pre-built according to the MBSE architecture standard;

[0083] The view standard specification requires that it at least include the metamodel specification defined in the OMG SysML, DoDAF, or UAF standard for the corresponding view; among which, the view standard specification requires the inclusion of the metamodel specification: the metamodel specification is the most core and fundamental rule in the view standard that must be followed, and the view standard specification also includes other requirements such as layout, format, and dependencies.

[0084] The business dependencies between views refer to the constraints between views defined using logical expressions or SPARQL statements.

[0085] The view rule system adopts JSON-LD and XML Schema standardized definition, and is stored in the form of modular JSON files, YAML files or graph databases. It supports calls via RESTful interfaces and supports dynamic addition, modification and version expansion. Acquisition methods include but are not limited to pre-built, real-time generated, cloud pull, third-party interface calls, and user-defined import.

[0086] Furthermore, the standardized modeling feature data is matched and analyzed with the view rule system, specifically including the following steps:

[0087] S31, Map the standardized modeling feature data to the view rule system using feature vectorization to obtain feature vectors and rule vectors, calculate the matching degree between feature vectors and rule vectors, and filter the views whose matching degree reaches the preset threshold to obtain a set of candidate views;

[0088] It needs to be explained that feature vectorization mapping specifically refers to mapping standardized modeling feature data into an n-dimensional feature vector, where n is the number of preset fields for the modeling elements, and mapping each view in the view rule system into an n-dimensional rule vector of the same dimension. Simultaneously, mandatory constraint terms are extracted from the rule terms. These mandatory constraint terms are constraints marked as mandatory conditions in the rule terms, and each mandatory constraint term is associated with a specific dimension of the feature vector, for a total of m mandatory constraint terms.

[0089] S32, perform constraint conflict verification and MBSE standard compliance verification on the candidate view set, remove conflicting and non-compliant views, obtain the types and number of views to be generated that match the target long text document, and automatically plan the view generation sequence based on the business topology dependencies between views to obtain the target view set.

[0090] Furthermore, the matching degree is calculated using a bidirectional matching algorithm, the specific steps of which include:

[0091] S311, Calculate the positive matching score between the feature vector and the rule vector. The positive matching score is used to characterize the overall similarity between the feature vector and the rule vector. The formula for calculating the positive matching score is:

[0092] ;

[0093] in, For positive matching scores, This refers to the number of dimensions in the feature vector, i.e., the number of preset fields for the modeling features. The summation index variable represents the i-th dimension of the feature vector. The normalized quantized value of the feature vector in the i-th dimension. Let be the normalized weight value of the rule vector in the i-th dimension;

[0094] S312, calculate the back-matching score between the mandatory constraint vector in the rule term and the associated dimension subset of the corresponding mandatory constraint in the feature vector. The back-matching score is used to characterize the completeness of the mandatory constraint requirements in the rule term being covered by the standardized modeling feature data. The formula for calculating the back-matching score is:

[0095] ;

[0096] in, For reverse matching scores, This represents the total number of dimensions for mandatory constraints in the current view's rule items, where j is the dimension index of the mandatory constraint item. Let J be the normalized weight of the j-th dimension mandatory constraint term in the rule term. This is the normalized quantized value of the dimension associated with the j-th dimension of the mandatory constraint term in the feature vector;

[0097] S313, the forward matching score and the backward matching score are weighted and summed to obtain the total matching degree. ,in, The weights for positive matching scores, The weights for the reverse matching score, ;

[0098] S314, filter the views whose total matching degree reaches the preset threshold to obtain a set of candidate views.

[0099] The preset threshold can be a dynamic threshold or a fixed threshold.

[0100] The dynamic threshold is the average matching degree of the candidate views, and the fixed threshold is 0.6.

[0101] Furthermore, the constraint conflict verification and MBSE standard compliance verification of the candidate view set includes the following steps:

[0102] S321, Extract business constraints and performance constraints from the target long text document;

[0103] S322, Constraint Conflict Validation: The extracted business constraints and performance constraints are compared with the rule requirements of the corresponding candidate views in the view rule system. Combined with the predefined business topology dependencies between views in the view rule system, conflicting views with logical mutual exclusion or missing dependencies in the candidate view set are determined and eliminated.

[0104] It should be explained that view rule requirements refer to the conditions that a view must meet, as specified in the Constraint (business dependencies between views) and Norm (view standard specification requirements) of the rule item quadruple. For example, "OV-1 view must depend on CV-1 view" is a view rule requirement.

[0105] S323, MBSE Standard Compliance Verification: Based on the mandatory specifications of the MBSE architecture standard, remove views with standard violations from the candidate view set;

[0106] It's important to clarify that the mandatory specifications of the MBSE architecture standard refer to the Norm (view standard specification requirements) in the rule item quadruple, as well as the mandatory rules defined in the entire MBSE architecture standard (such as DoDAF and UAF) that must be followed. These include specific elements that a view must contain (e.g., SV-1 views must contain system interfaces), the legal relationship types between elements, the range of attribute values ​​or data types, and the naming rules for views. These specifications are typically marked as "mandatory" in the rule items and are explicitly stated in the standard documents.

[0107] S324, After both constraint conflict verification and standard compliance verification are completed, the remaining views in the candidate view set are sorted from highest to lowest matching degree, and the top N views with the highest matching degree are retained. The value of N is determined by rounding down from 20% of the total number of candidate views to a range of 3-8. Therefore, the number of retained candidate views is... M is the total number of candidate views after verification;

[0108] Unlike traditional single-view generation which requires manual specification of view types, this invention can automatically complete view selection, quantity determination, and sequence planning, with high view matching accuracy and no need for manual intervention.

[0109] S325, combining the business topology dependencies between views, automatically plans the view generation sequence to obtain the target view set.

[0110] In this embodiment, the matching degree is obtained by means of at least one of semantic vectorization matching, rule engine triggering, and machine learning classification.

[0111] Firstly, the specific steps for calculating the matching degree using the semantic vectorization matching method include:

[0112] By using a pre-trained language model, text-based modeling features in standardized modeling feature data are encoded into fixed-dimensional modeling feature semantic vectors.

[0113] The rule items in the view rule system are encoded into view rule semantic vectors with the same fixed dimension, and the value range of the fixed dimension is [128, 768].

[0114] The matching degree between the modeling feature semantic vector and the view rule semantic vector is calculated using the cosine similarity formula. The matching degree calculation formula based on the cosine similarity formula is as follows:

[0115] ;

[0116] in, The matching degree is calculated based on the cosine similarity formula, where k is the total number of fixed dimensions of the semantic vectors, and r is the index of the fixed dimension of the semantic vectors. To model the numerical values ​​of the feature semantic vector in the r-th dimension, Let r be the numerical value of the view rule semantic vector in the r-th dimension;

[0117] Filter the views that meet the matching criteria to form a candidate view set.

[0118] Secondly, the specific steps for calculating the matching degree using the rule engine triggering method include:

[0119] The view rule system is encapsulated into a rule engine library, where each rule corresponds to the view generation trigger conditions, constraint requirements and output specifications of a view;

[0120] Standardized modeling feature data is fed into the rule engine library as input data. The rule engine library matches the modeling features with the view to generate trigger conditions through condition triggering logic, and filters the views that meet all trigger conditions as a set of candidate views.

[0121] Third, the specific steps for calculating the matching degree using machine learning classification include:

[0122] Based on historical equipment modeling data, sample annotation was completed to construct a training dataset;

[0123] The view classification model is trained using the training dataset to obtain a pre-trained view classification model, which is used to output the corresponding view type based on the input modeling features.

[0124] Standardized modeling feature data is input into the pre-trained view classification model, which directly outputs a set of suitable candidate views.

[0125] Furthermore, the specific steps of the batch multi-view modeling include:

[0126] S411, decompose the target view set in the task dimension to obtain distributed modeling subtasks corresponding to multiple single view generation tasks;

[0127] S412, the distributed modeling subtask is executed through multi-threaded parallel processing, with each thread completing a modeling task for a single view data item and generating the single view data items;

[0128] It should be explained that this invention adopts a multi-agent cluster architecture, setting up a master control and scheduling agent and dedicated functional agents, including a design scheme parsing agent, a view matching agent, a view generation agent, and a consistency verification agent; each agent completes data interaction and state synchronization through a message middleware using JSON messages.

[0129] The master control agent breaks down the core task into sub-tasks and assigns them to corresponding dedicated agents for distributed collaborative execution. It is also responsible for task scheduling, status monitoring, and exception handling. Each dedicated agent processes data in parallel and shares modeling data, avoiding redundant calculations and data conflicts, and improving the efficiency of large-scale batch modeling execution.

[0130] S413 performs a three-level cross-view feature consistency check on each individual view data at the feature level, view level, and system level. If the consistency check passes, the individual view data is directly output as the corrected view feature. The check content includes the consistency of attribute values ​​of the same modeling feature in different views, as well as the compliance of the individual view data with the MBSE system architecture standard metamodel specification.

[0131] It should be explained that this invention pre-constructs a directed constraint relationship graph between views, clearly defining the pre-dependencies and post-dependencies of each view, forming a directed acyclic graph (DAG) of view generation dependencies. Based on the topology sorting algorithm, combined with the target view set, the optimal execution logic is determined. Dependency-free views are generated in parallel, while dependent views are executed serially in a pipeline, thus balancing modeling efficiency and cross-view compatibility. Figure 1 To the point of being responsive.

[0132] S414, when a single view data fails the consistency check, a baseline view is determined according to the preset priority rules, and the attribute values ​​of the corresponding inconsistent elements in other views are uniformly corrected to the attribute values ​​of the element in the baseline view to obtain the corrected view features; the preset priority rules are determined in the following order: the first priority is the preceding root view in the business topology dependency relationship between views, the second priority is the view with the highest mandatory specification level in the MBSE system architecture standard, and the third priority is the view with the highest total matching degree.

[0133] S415, the corrected view features are mapped element-by-element to the preset view visualization parameters to complete the visualization rendering and obtain a multi-view model. The preset view visualization parameters are view visualization specifications adapted to the corresponding MBSE architecture standard, including standard layout rules for individual views, visualization style definitions for modeling elements, and mapping rules between metamodels and graphic elements.

[0134] Furthermore, the cross-view feature consistency verification of each individual view data includes: performing three-level quantitative verification of the individual view data at the feature level, view level, and system level, that is, checking whether the attribute values ​​of the same modeling feature are consistent in different views, and checking whether each individual view data conforms to the metamodel specification of the MBSE system architecture standard. Specific steps include:

[0135] S4131, For the numerical modeling features of each individual view data in the multi-view model, calculate the feature-level deviation D. The formula for calculating the feature-level deviation D is: ,in , Attribute values ​​of the same numerical modeling feature in different views. The maximum permissible attribute deviation value preset for this modeling element in the MBSE system architecture standard is used for deviation determination in consistency verification. The value range of D is [0,1].

[0136] S4132, For the number of rule items in each individual view data in the multi-view model, calculate the view-level compliance score C. The formula for calculating the view-level compliance score C is as follows: ,in This represents the number of rule items in the single-view data that conform to the MBSE standard specification. The total number of rule items in this single view data, C, has a value range of [0,1].

[0137] S4133, Based on the business topology dependency relationship between views, obtain the preset view topology association weight coefficient W for each individual view data, and the sum of the view topology association weight coefficients of all views is 1.

[0138] S4134, For each individual view data item, calculate the consistency evaluation index Q. ,in, , , These are the weighting coefficients corresponding to feature-level deviation, view-level compliance score, and preset view topology association weighting coefficients, respectively. , , All are positive numbers, and ;

[0139] It needs to be explained that, , , The method for determining the maximum number of iterations is as follows: it is calculated based on the number of modeling iterations and the preset maximum number of iterations, with the preset maximum number of iterations ranging from [10, 20].

[0140] S4135, Obtain the preset convergence threshold. When consistency evaluation indicators At that time, it is determined that the single-view data passes the consistency check, and the modeling iteration converges. If the data in a single view fails the consistency check, step S414 is triggered.

[0141] It needs to be explained that the preset convergence threshold... The value range is [0.85, 0.95].

[0142] Example

[0143] DoDAF system modeling and implementation for military equipment scenarios:

[0144] This embodiment targets the construction scenario of military equipment systems. Based on the DoDAF 2.0 standard, it automatically generates multi-view models, supplementing specific parameters, intermediate data, and operational details. The specific implementation process is as follows:

[0145] Input: A user uploaded a DOCX document (2.8MB, 12 pages) outlining a design scheme for a certain type of army equipment system. The core content is as follows:

[0146] "The goal of this equipment system is to achieve integrated coordination of command, reconnaissance, firepower, and support. The organizational structure includes the army group command center, armored brigade reconnaissance battalion, firepower strike regiment, and logistics support brigade. The combat missions include deep strike, intelligence reconnaissance, firepower coverage, and logistics support. The system consists of command and control unit (CCU), reconnaissance and early warning unit (REU), firepower strike unit (FSU), and logistics support unit (LOG). The data interaction requirements are that the REU transmits target coordinates to the CCU in real time, the CCU issues firepower strike commands to the FSU, and the LOG unit shares material status data with each combat unit. The capability indicators are target positioning error ≤50m, firepower response time ≤10s, and data transmission latency ≤200ms."

[0147] Processing flow:

[0148] S1: Obtain the pre-built DoDAF 2.0 standard view rule system. This rule system adopts a four-level architecture and JSON-LD standardized definition. The core rule items are shown in the following example (excerpt):

[0149] DoDAF2.0 CV (Vision) CV-1 {Sim≥0.85,{"Equipment System Objective","Integrated Collaboration"},"SysML::BlockDefinition","No Cross-View Constraints"} DoDAF2.0 OV (Operation) OV-1 {Sim≥0.80,{"Operational Mission","Organizational Structure"},"DoDAF::OV-1::OperationalNode","Depends on CV-1 Vision View"} DoDAF2.0 SV (System) SV-1 {Sim≥0.80,{"System Composition","Data Interaction"},"DoDAF::SV-1::SystemInterface","Dependency OV-5a Activity Model View"}

[0150] S2: Perform structured parsing on the input DOCX document, using a dual-mode parsing approach of "large language model + military domain knowledge graph". The intermediate parsing results are as follows:

[0151] Extract core modeling elements: organizational nodes (army group command center, armored brigade reconnaissance battalion), combat missions (deep strike, intelligence reconnaissance), system components (CCU, REU, FSU, LOG), data links (target coordinate transmission, fire command issuance), and capability indicators (positioning error ≤50m).

[0152] Output standardized MBSE modeling feature data: including entity vectors (e.g., [0.92, 0.85, 0.78, 0.90, 0.88], corresponding to the semantics of "command-reconnaissance-firepower-support-integration"), entity element sets {organization nodes, combat missions, system components, data links}, and business constraints {data transmission latency ≤ 200ms}.

[0153] S3: Based on the DoDAF 2.0 view rule system, it performs bidirectional matching analysis. The specific calculation and matching process is as follows:

[0154] Feature vectorization: The parsed modeling features are converted into BERT embedding vectors V_feature, and the view rule keywords are converted into pre-trained word vectors V_rule;

[0155] Forward matching: Calculate the cosine similarity Sim=cos(V_feature,V_rule), and combine it with TF-IDF to weight core features such as "command and control" and "firepower strike" (weight coefficient 0.2) to obtain the weighted matching score;

[0156] Threshold determination: Set the matching degree threshold to 0.80, and filter out the candidate view set {CV-1,CV-2,OV-1,OV-2,OV-4,OV-5a,SV-1,SV-4} with a matching degree ≥80%;

[0157] Reverse validation: Based on the constraint relationships of the rule items, empty views that do not meet the constraint of "data transmission delay ≤ 200ms" are removed, and the final target view set is determined. The output view list is as follows:

[0158] CV-1 DoDAF2.0 0.92 1 none CV-2 DoDAF2.0 0.89 2 Depends on CV-1 OV-1 DoDAF2.0 0.88 3 Depends on CV-1 OV-2 DoDAF2.0 0.85 4 Depends on OV-1 OV-4 DoDAF2.0 0.83 5 Depends on OV-1 OV-5a DoDAF2.0 0.82 6 Dependent on OV-2 SV-1 DoDAF2.0 0.86 7 Dependent on OV-5a SV-4 DoDAF2.0 0.81 8 Depends on SV-1

[0159] Specifically, a Directed Acyclic Graph (DAG) can be constructed based on the constraint dependency graph between DoDAF views. The specific dependencies and scheduling logic are as follows:

[0160] Directed constraint relations: CV-1→CV-2→OV-1→OV-2→OV-5a→SV-1→SV-4, OV-1→OV-4;

[0161] Scheduling strategy: Undependent OV-4 and CV-2 are generated in parallel, while views with prerequisite dependencies are generated serially in the order of dependencies, forming an execution chain of "CV-1 serial → {CV-2, OV-1} parallel → OV-2 serial → OV-5a serial → SV-1 serial → SV-4 serial".

[0162] S4: According to the scheduling logic, call the specification requirements in the DoDAF 2.0 view rule system to complete the automatic generation of multiple views. The generation time of a single view is: CV-1 (12s), CV-2 (15s), OV-1 (18s), OV-2 (20s), OV-4 (16s), OV-5a (22s), SV-1 (25s), SV-4 (20s), with a total generation time of 95s.

[0163] After generation, a three-level consistency quantification check was strictly performed at the element level, view level, and system level. Cross-view element consistency check was also performed. The specific check indicators, calculation process, and measured results are as follows:

[0164] Element-level verification: For core numerical modeling elements such as organizational nodes and combat missions, the cross-view element-level deviation degree D is calculated. The maximum value of the attribute deviation of the core elements is 0.04, corresponding to a core element entity overlap rate Recall=96%, which is better than the preset ≥90% qualified threshold. There are no missing elements or deviations exceeding the limit.

[0165] View-level validation: By traversing the AST abstract syntax tree, it is verified that all individual view data conforms to the DoDAF2.0 metamodel specification. The percentage of rule items in the individual view that conform to the standard specification is 100%, that is, the view-level compliance score C=1.0, with no standard violations or syntax errors.

[0166] System-level validation: Based on the business topology dependencies between views, a topology association weight coefficient W is preset for each view, and the sum of all weight coefficients is 1; combined with the preset weight coefficients... =0.3、 =0.4、 =0.3, the final consistency evaluation index Q=0.97 was calculated, which is better than the preset convergence judgment threshold of ≥0.8, and all views were judged to have passed the consistency verification; after the verification was completed, a compliance verification report was output. The report included the measured values ​​of each level of verification, deviation details and compliance judgment results. There were no abnormal prompts in this verification.

[0167] Output a compliance verification report, including verification values ​​at each level and any error messages.

[0168] S5: Convert the verified multi-view model to XMI format to generate a standardized model file (file size 4.2MB), which can be directly imported into MagicDraw 19.0 for secondary editing.

[0169] Output results: 8 core multi-view models conforming to the DoDAF 2.0 standard, visual view files, and compliance verification report; according to actual tests, the modeling cycle of this embodiment is 12 minutes (including parsing, generation, and verification), which is 90% more efficient than traditional manual modeling (120 minutes), with a 100% view compliance rate. Even those with no prior experience can complete the modeling without intervention simply by uploading the document.

[0170] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0171] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0172] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0173] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0174] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0175] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for automatically generating multi-view models adapted to the MBSE architecture standard, characterized in that, Includes the following steps: S1. Obtain the view rule system that adapts to the MBSE architecture standard. The view rule system is a set of view rules that includes view type characteristics, view generation trigger conditions, view standard specification requirements, and business topology dependencies between views. S2 performs structured parsing on the input target long text document, extracting modeling elements, business constraints, and design requirements from the target long text document to obtain standardized modeling feature data; S3, by matching and analyzing the standardized modeling feature data with the view rule system, the types and number of views to be generated that match the target long text document are automatically determined, and the view generation sequence is automatically planned in combination with the business topology dependency relationship between views to obtain the target view set; S4. By calling the view standard specifications in the view rule system, batch multi-view modeling is automatically performed according to the target view set, and conflicts are corrected through cross-view element consistency verification to obtain the multi-view model. S5. Perform standardized format conversion on the multi-view model and output a standardized model file that conforms to the MBSE modeling tool compatible format.

2. The method for automatically generating multi-view models adapted to the MBSE architecture standard as described in claim 1, characterized in that, The view rule system adopts a four-level extensible architecture, including a standard version layer, a view group layer, a view layer, and a rule item layer; The standard version layer is the adapted MBSE architecture version number, used to adapt to different versions of MBSE standard specifications; The view group layer is a collection of views divided by business domain, used to classify and manage views within the same business domain; The view layer is a specific MBSE single-view data type, corresponding to the smallest unit of single-view data modeling; Each rule item in the rule item layer is defined using a four-tuple structure. The four-tuple includes the view generation trigger condition, feature keywords, view standard specification requirements, and business topology dependencies between views, which are used to define the view generation rules.

3. The method for automatically generating multi-view models adapted to the MBSE architecture standard as described in claim 2, characterized in that, The standardized modeling feature data is matched and analyzed with the view rule system, specifically including the following steps: S31, Map the standardized modeling feature data to the view rule system using feature vectorization to obtain feature vectors and rule vectors, calculate the matching degree between feature vectors and rule vectors, and filter the views whose matching degree reaches the preset threshold to obtain a set of candidate views; S32, perform constraint conflict verification and MBSE standard compliance verification on the candidate view set, remove conflicting and non-compliant views, obtain the types and number of views to be generated that match the target long text document, and automatically plan the view generation sequence based on the business topology dependencies between views to obtain the target view set.

4. The method for automatically generating multi-view models adapted to the MBSE architecture standard as described in claim 3, characterized in that, The matching degree is calculated using a bidirectional matching algorithm, and the specific steps include: S311, Calculate the positive matching score between the feature vector and the rule vector. The positive matching score is used to characterize the overall similarity between the feature vector and the rule vector. The formula for calculating the positive matching score is: ； in, For positive matching scores, This refers to the number of dimensions in the feature vector, i.e., the number of preset fields for the modeling features. The summation index variable represents the i-th dimension of the feature vector. The normalized quantized value of the feature vector in the i-th dimension. Let be the normalized weight value of the rule vector in the i-th dimension; S312, calculate the back-matching score between the mandatory constraint vector in the rule term and the associated dimension subset of the corresponding mandatory constraint in the feature vector. The back-matching score is used to characterize the completeness of the mandatory constraint requirements in the rule term being covered by the standardized modeling feature data. The formula for calculating the back-matching score is: ； in, For reverse matching scores, This represents the total number of dimensions for mandatory constraints in the current view's rule items, where j is the dimension index of the mandatory constraint item. Let J be the normalized weight of the j-th dimension mandatory constraint term in the rule term. This is the normalized quantized value of the dimension associated with the j-th dimension of the mandatory constraint term in the feature vector; S313, the forward matching score and the backward matching score are weighted and summed to obtain the total matching degree. ,in, The weights for positive matching scores, The weights for the reverse matching score, ; S314, filter the views whose total matching degree reaches the preset threshold to obtain a set of candidate views.

5. The method for automatically generating multi-view models adapted to the MBSE architecture standard as described in claim 3, characterized in that, The constraint conflict verification and MBSE standard compliance verification of the candidate view set includes the following steps: S321, Extract business constraints and performance constraints from the target long text document; S322, Constraint Conflict Validation: The extracted business constraints and performance constraints are compared with the rule requirements of the corresponding candidate views in the view rule system. Combined with the predefined business topology dependencies between views in the view rule system, conflicting views with logical mutual exclusion or missing dependencies in the candidate view set are determined and eliminated. S323, MBSE Standard Compliance Verification: Based on the mandatory specifications of the MBSE architecture standard, remove views with standard violations from the candidate view set; S324. After both constraint conflict verification and standard compliance verification are completed, sort the remaining views in the candidate view set from high to low matching degree, and retain the top N views with the highest matching degree. S325, combining the business topology dependencies between views, automatically plans the view generation sequence to obtain the target view set.

6. The method for automatically generating multi-view models adapted to the MBSE architecture standard as described in claim 1, characterized in that, The specific steps of batch multi-view modeling include: S411, decompose the target view set in the task dimension to obtain distributed modeling subtasks corresponding to multiple single view generation tasks; S412, the distributed modeling subtask is executed through multi-threaded parallel processing, with each thread completing a modeling task for a single view data item and generating the single view data items; S413 performs a three-level cross-view feature consistency check on each individual view data at the feature level, view level, and system level. If the consistency check passes, the individual view data is directly output as the corrected view feature. S414 When the data of a single view fails the consistency check, a baseline view is determined according to the preset priority rules. The attribute values ​​of the corresponding inconsistent elements in other views are uniformly corrected to the attribute values ​​of the element in the baseline view to obtain the corrected view features. S415 maps the corrected view features to preset view visualization parameters element by element to complete the visualization rendering and obtain a multi-view model.

7. The method for automatically generating multi-view models adapted to the MBSE architecture standard as described in claim 6, characterized in that, The cross-view feature consistency verification of individual view data includes: checking whether the attribute values ​​of the same modeling feature are consistent in different views, and checking whether the individual view data conforms to the metamodel specification of the MBSE architecture standard. Specific steps include: S4131, For the numerical modeling features of each individual view data in the multi-view model, calculate the feature-level deviation D. The formula for calculating the feature-level deviation D is: ,in , Attribute values ​​of the same numerical modeling feature in different views. The maximum permissible attribute deviation value preset for this modeling element in the MBSE system architecture standard is used for deviation determination in consistency verification. The value range of D is [0,1]. S4132, For the number of rule items in each individual view data in the multi-view model, calculate the view-level compliance score C. The formula for calculating the view-level compliance score C is as follows: ,in This represents the number of rule items in the single-view data that conform to the MBSE standard specification. The total number of rule items in this single view data, C, has a value range of [0,1]. S4133, Based on the business topology dependency relationship between views, obtain the preset view topology association weight coefficient W for each individual view data, and the sum of the view topology association weight coefficients of all views is 1; S4134, For each individual view data item, calculate the consistency evaluation index Q. ,in, , , These are the weighting coefficients corresponding to feature-level deviation, view-level compliance score, and preset view topology association weighting coefficients, respectively. , , All are positive numbers, and ; S4135, Obtain the preset convergence threshold. When consistency evaluation indicators When the data in a single view passes the consistency check, it is determined that the data in that single view passes the consistency check. If the data in a single view fails the consistency check, step S414 is triggered.

Images