Intelligent meta-model construction and dynamic adaptation method and system for ship field

Abstract

The invention discloses a ship field-oriented intelligent meta-model construction and dynamic adaptation method and system. The method comprises the following steps: constructing a ship field system meta-model library comprising concept, interaction and behavior meta-models; performing semantic understanding and intention recognition on natural language requirements of a user by using a large language model of an integrated domain dictionary, and extracting structured modeling elements; word vector cosine similarity and semantic distance calculation based on domain ontology are fused to generate a candidate set; a multi-objective decision model is adopted, semantic fitting degree, coupling degree and multiplexing frequency are comprehensively considered to optimize and screen the candidate set, and an optimal meta-model adaptation sequence is output; instantiating the meta-model and constructing an incidence relation; performing consistency verification including static grammar, dynamic semantics and cross-view logic on the instance set; and performing version persistent storage on the verified model. According to the method, automation and intelligence of ship domain model construction are realized, and the modeling efficiency and the model accuracy and reliability are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of ship design engineering, and in particular relates to a method and system for constructing and dynamically adapting intelligent meta-models for the shipbuilding field. Background Technology

[0002] In the fields of ship design, manufacturing, and operation and maintenance, building accurate and efficient domain models is the core foundation for achieving digitalization and intelligence. Domain models, through logical abstractions of ship entities, interactive behaviors, and state transitions, provide crucial support for simulation verification, system integration, and decision support. Traditional methods for building ship domain models heavily rely on human experience. Modelers must manually define model elements, attributes, and relationships—a process that is not only time-consuming and labor-intensive but also highly susceptible to subjective errors, resulting in inconsistent model quality. Especially when facing complex natural language requirements, the ambiguity of human interpretation further reduces the model's relevance to real-world business needs.

[0003] Existing technologies include several tools and methods that attempt to improve modeling efficiency. For example, template-based code generation or semi-automatic modeling using standardized modeling languages ​​such as UML. However, these methods often have significant limitations: Insufficient flexibility: Templates and fixed syntax are difficult to cover the diverse business scenarios in the shipping industry (such as the unique operating procedures of special vessels) and cannot adapt to the modeling intentions of different users.

[0004] Lack of semantic understanding: Existing methods lack the ability to deeply understand the semantic requirements of natural language and cannot effectively extract complex constraints such as "collision avoidance rules" and "multi-ship collaborative channel scheduling", resulting in deviations in the behavioral logic of the generated models.

[0005] Weak Model Consistency Guarantee: Ship system models typically need to be described from multiple perspectives, including structure, behavior, and interaction. Existing tools do not adequately support dynamic logical consistency verification between models of different perspectives, relying heavily on manual review, which makes it difficult to effectively detect deep-seated problems such as "state deadlock" and "rule conflicts".

[0006] Poor reusability: Although some meta-model libraries exist, they lack intelligent matching and recommendation mechanisms. Modelers struggle to quickly and accurately locate and reuse the most suitable components from massive meta-model datasets, leading to widespread duplication of development and wasted resources.

[0007] Therefore, there is an urgent need in this field for an automated method that can intelligently understand user intent, automatically match and optimize meta-model combinations, and perform rigorous multi-dimensional consistency verification on the generated model, so as to fundamentally improve the efficiency, accuracy and reliability of model building in the shipbuilding field. Summary of the Invention

[0008] To address the shortcomings of the existing technologies, this invention provides a method for constructing and dynamically adapting an intelligent meta-model for the shipbuilding field, comprising the following steps: S1. Construct a meta-model library for shipbuilding systems, which includes conceptual meta-models, interaction meta-models, and behavioral meta-models. S2. Utilize a large language model to perform semantic understanding and modeling intent recognition on the natural language input from users, extract modeling elements, and form a structured set of modeling elements. S3. Based on the set of structured modeling elements obtained in step S2, perform meta-model matching in the meta-model library through semantic similarity calculation to generate a set of candidate meta-models; S4. Perform multi-criteria optimization and screening on the candidate meta-model set generated in step S3, and output the optimal meta-model adaptation sequence. S5. Based on the optimal metamodel adaptation sequence obtained in step S4, instantiate metamodel instances and construct the association mapping relationship between metamodel instances. S6. Perform a consistency check on the set of meta-model instances generated in step S5; S7. Persistently store the verified meta-model instances and their relationships in the ship domain model library, and generate corresponding model documents and configuration management information.

[0009] In step S1, the conceptual meta-model is used to define entities and their static attributes within the ship domain, the interaction meta-model is used to define dynamic interaction relationships and protocols between entities, and the behavioral meta-model is used to define the state transition logic and behavioral rules of entities; and the construction of the meta-model library is based on historical AIS data, ship navigation logs, and domain ontology.

[0010] In step S1, the conceptual meta-model includes ship, port, and waterway entities; the interaction meta-model includes ship-port communication protocol; and the behavioral meta-model includes ship navigation state change rules.

[0011] In step S2, the large language model is trained based on the BERT architecture and integrates a marine domain dictionary; the extracted modeling elements include domain entity keywords, interaction relationship descriptions, and behavioral constraints.

[0012] In step S3, the semantic similarity calculation integrates word vector cosine similarity with semantic distance based on the ship domain ontology.

[0013] In step S4, the multi-criteria optimization screening comprehensively considers the semantic fit between the metamodel and user needs, the coupling degree between metamodels, and the reuse frequency of the metamodel in historical modeling projects; and is achieved by solving a multi-objective optimization problem, the objective function of which aims to maximize the overall fit utility value U: ; Where n is the number of candidate meta-models, Si is the semantic fit, Ci is the coupling degree, Ri is the normalized reuse frequency, and α, β, and γ are weight coefficients dynamically determined by the AHP (Analog-Philosophy of Things) method.

[0014] The process includes step S5a after step S5 and before step S6: performing a complexity assessment on the initially instantiated metamodel. If the assessment value exceeds a preset threshold, a metamodel simplification operation is triggered. The simplification operation includes merging metamodel instances with similar functions, removing redundant attributes, or abstracting low-level behaviors. The complexity is calculated based on the number of nodes and relations of the metamodel instance.

[0015] In step S6, the consistency verification includes static syntax verification, dynamic semantic verification, and cross-view logic consistency verification. The dynamic semantic verification includes: verifying the state transition path based on linear temporal logic formulas, and automatically traversing the behavior trajectory using model detection technology to check whether there is a violation of maritime safety regulations.

[0016] In step S7, the persistent storage adopts a versioned storage strategy, generating a new model version each time it is stored and recording the differences between versions; the configuration management information includes the creator of the model element, the creation time, the last modification time, and the dependencies.

[0017] This invention also proposes an intelligent meta-model construction and dynamic adaptation system for the shipbuilding field based on the above method, comprising: The meta-model library management module is used to store and manage the meta-model library of the ship domain system. The natural language processing module is used to perform semantic understanding and modeling intent recognition in step S2; The meta-model matching and recommendation module is used to perform semantic similarity calculation and meta-model matching in step S3; The optimization filtering module is used to perform multi-criteria optimization filtering in step S4. The instantiation module is used to perform the metamodel instantiation and association mapping in step S5; The consistency verification module is used to perform the consistency verification in step S6; The storage management module is used to perform persistent storage and version management in step S7.

[0018] Compared with the prior art, the present invention has the following advantages: This method significantly improves modeling efficiency and automation. By automatically parsing natural language requirements and extracting key elements through a large language model, it changes the traditional model that relies entirely on manual interpretation and modeling. Combined with intelligent meta-model matching, optimization filtering, and instantiation processes, it frees modelers from tedious and repetitive tasks, automating and intelligently building the model. To verify the effectiveness of this method, a comparative experiment was conducted. Using traditional manual modeling as a baseline, this method was used to model 10 typical ship scheduling scenarios. The results show that the average modeling time of this method is reduced from 40 person-hours to 2 person-hours, improving modeling efficiency by approximately 95%. Formal verification of the generated models revealed that the rate of behavioral logic violations of domain safety regulations in the models generated by this method is 80% lower than that of the baseline model.

[0019] Significantly improves model accuracy and demand relevance. Employing a semantic similarity calculation method that integrates word vector cosine similarity and ontology semantic distance, compared to traditional keyword matching, it can more deeply understand the semantic connotations of user needs, thus obtaining more accurate candidate meta-models that better match the user's true intent during the meta-model matching stage. Multi-criteria optimization further ensures high semantic relevance of the final adapted sequence from a global perspective.

[0020] It effectively ensures the internal consistency and correctness of the model. It innovatively introduces a multi-granularity, multi-view consistency verification mechanism, particularly dynamic semantic verification based on Linear Temporal Logic (LTL) formulas and model checking technology. This mechanism can automatically traverse the behavioral trajectories of meta-model instances, efficiently detecting potential state deadlocks, rule conflicts, and security specification violations that are difficult for humans to detect. This significantly improves the internal consistency and behavioral correctness of the generated model, reducing the system development risks and costs caused by model errors in later stages.

[0021] Enhanced model reusability and system maintainability. By establishing a rich and standardized metamodel library and introducing optimization criteria based on historical reuse frequency, the reuse of validated, high-quality metamodels is encouraged. This not only reduces redundant development but also improves the quality and reliability of newly constructed models. The versioned persistent storage strategy clearly records the model's evolution history and dependencies, greatly facilitating subsequent model maintenance, upgrades, and traceability.

[0022] It exhibits good domain adaptability and scalability. The core framework and key technologies of the method (such as semantic computation, optimization screening, and consistency verification) are not limited to the shipbuilding field. By replacing domain-related ontology libraries, metamodel libraries, and rule bases, it can be easily adapted to other complex equipment manufacturing, transportation, and other fields, demonstrating good versatility and scalability prospects.

[0023] In summary, this invention effectively solves the long-standing pain points in the model building process of the shipbuilding industry, such as low efficiency, insufficient accuracy, and difficulty in ensuring consistency, and provides solid and reliable technical support for the digital transformation and intelligent upgrading of the shipbuilding industry. Attached Figure Description

[0024] The above and other objects, features, and advantages of exemplary embodiments of the present disclosure will become readily apparent upon reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the present disclosure are illustrated by way of example and not limitation, and like or corresponding reference numerals denote like or corresponding parts, wherein: Figure 1 This is a flowchart illustrating a method for constructing and dynamically adapting an intelligent meta-model for the shipbuilding field according to an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0026] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0027] It should be understood that although the terms first, second, third, etc., may be used to describe... in the embodiments of the present invention, these... should not be limited to these terms. These terms are only used to distinguish... For example, first... may also be referred to as second... without departing from the scope of the embodiments of the present invention, and similarly, second... may also be referred to as first...

[0028] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0029] Depending on the context, the words “if” or “suppose” as used here can be interpreted as “when” or “in response to determination” or “in response to detection.” Similarly, depending on the context, the phrases “if determination” or “if detection (of the stated condition or event)” can be interpreted as “when determination” or “in response to determination” or “when detection (of the stated condition or event)” or “in response to detection (of the stated condition or event).”

[0030] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that an article or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device that includes said element.

[0031] The optional embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0032] Example 1 like Figure 1 As shown, this invention discloses a method for constructing and dynamically adapting an intelligent meta-model for the shipbuilding field, comprising the following steps: S1. Construct a meta-model library for shipbuilding systems, which includes conceptual meta-models, interaction meta-models, and behavioral meta-models. S2. Utilize a large language model to perform semantic understanding and modeling intent recognition on the natural language input from users, extract modeling elements, and form a structured set of modeling elements. S3. Based on the set of structured modeling elements obtained in step S2, perform meta-model matching in the meta-model library through semantic similarity calculation to generate a set of candidate meta-models; S4. Perform multi-criteria optimization and screening on the candidate meta-model set generated in step S3, and output the optimal meta-model adaptation sequence. S5. Based on the optimal metamodel adaptation sequence obtained in step S4, instantiate metamodel instances and construct the association mapping relationship between metamodel instances. S6. Perform a consistency check on the set of meta-model instances generated in step S5; S7. Persistently store the verified meta-model instances and their relationships in the ship domain model library, and generate corresponding model documents and configuration management information.

[0033] In step S1, the conceptual meta-model is used to define entities and their static attributes within the ship domain, the interaction meta-model is used to define dynamic interaction relationships and protocols between entities, and the behavioral meta-model is used to define the state transition logic and behavioral rules of entities; and the construction of the meta-model library is based on historical AIS data, ship navigation logs, and domain ontology.

[0034] In step S1, the conceptual meta-model includes ship, port, and waterway entities; the interaction meta-model includes ship-port communication protocol; and the behavioral meta-model includes ship navigation state change rules.

[0035] In step S2, the large language model is trained based on the BERT architecture and integrates a marine domain dictionary; the extracted modeling elements include domain entity keywords, interaction relationship descriptions, and behavioral constraints.

[0036] In step S3, the semantic similarity calculation integrates word vector cosine similarity with semantic distance based on the ship domain ontology.

[0037] In step S4, the multi-criteria optimization screening comprehensively considers the semantic fit between the metamodel and user needs, the coupling degree between metamodels, and the reuse frequency of the metamodel in historical modeling projects; and is achieved by solving a multi-objective optimization problem, the objective function of which aims to maximize the overall fit utility value U: ; Where n is the number of candidate meta-models, Si is the semantic fit, Ci is the coupling degree, Ri is the normalized reuse frequency, and α, β, and γ are weight coefficients dynamically determined by the AHP (Analog-Philosophy of Things) method.

[0038] The process includes step S5a after step S5 and before step S6: performing a complexity assessment on the initially instantiated metamodel. If the assessment value exceeds a preset threshold, a metamodel simplification operation is triggered. The simplification operation includes merging metamodel instances with similar functions, removing redundant attributes, or abstracting low-level behaviors. The complexity is calculated based on the number of nodes and relations of the metamodel instance.

[0039] In step S6, the consistency verification includes static syntax verification, dynamic semantic verification, and cross-view logic consistency verification. The dynamic semantic verification includes: verifying the state transition path based on linear temporal logic formulas, and automatically traversing the behavior trajectory using model detection technology to check whether there is a violation of maritime safety regulations.

[0040] In step S7, the persistent storage adopts a versioned storage strategy, generating a new model version each time it is stored and recording the differences between versions; the configuration management information includes the creator of the model element, the creation time, the last modification time, and the dependencies.

[0041] Example 2 This invention proposes a method for constructing and dynamically adapting an intelligent meta-model for the shipbuilding field, comprising the following steps: S1. Construct a meta-model library for the ship domain system. The meta-model library includes a conceptual meta-model, an interaction meta-model, and a behavioral meta-model. The conceptual meta-model is used to define entities and their static attributes within the ship domain. The interaction meta-model is used to define the dynamic interaction relationships and protocols between entities. The behavioral meta-model is used to define the state transition logic and behavioral rules of entities. The construction of the meta-model library is based on historical AIS data, ship navigation logs, and domain ontology. The conceptual meta-model includes ship, port, and waterway entities. The interaction meta-model includes ship-port communication protocols. The behavioral meta-model includes ship navigation state transition rules. S2. Utilize a large language model to perform semantic understanding and modeling intent recognition on the natural language requirements input by the user, extracting domain entity keywords, interaction relationship descriptions, and behavioral constraints to form a set of structured modeling elements; the large language model is trained based on the BERT architecture and integrates a ship domain dictionary to improve recognition accuracy; S3. Based on the structured modeling element set obtained in step S2, perform meta-model matching in the ship domain system meta-model library through semantic similarity calculation to generate a candidate meta-model set. The semantic similarity calculation integrates word vector cosine similarity and semantic distance based on ship domain ontology. Specifically, word vector cosine similarity is calculated using word vectors trained by the Word2Vec model, and semantic distance based on ship domain ontology is obtained by calculating the shortest path distance of concepts in the ontology graph and weighting it by information content. S4. Perform multi-criteria optimization and screening on the candidate meta-model set generated in step S3. The multi-criteria optimization and screening uses a multi-objective decision model, which comprehensively considers the semantic fit between the meta-model and user needs, the coupling degree between meta-models, and the reuse frequency of meta-models in historical modeling projects, and outputs the optimal meta-model adaptation sequence. The multi-objective decision model adopts a weighted sum method, and the weights are dynamically adjusted according to the modeling context. S5. Based on the optimal metamodel adaptation sequence obtained in step S4, instantiate metamodel instances that conform to ship modeling specifications, and construct the association mapping relationship between metamodel instances; the instantiation process includes attribute assignment, relationship binding and rule activation, and generates an XML format model file; S6. Perform multi-granularity, multi-view consistency verification on the meta-model instance set generated in step S5. The consistency verification includes static syntax verification, dynamic semantic verification, and cross-view logic consistency verification to ensure the consistency and consistency of model elements across different abstraction levels and different views. Among them, dynamic semantic verification verifies the state transition path of behavioral meta-model instances based on linear temporal logic formulas and uses model detection tools such as NuSMV to automatically traverse possible behavioral trajectories. S7. Persistently store the verified meta-model instances and their relationships in the ship domain model library, and generate corresponding model documents and configuration management information; the persistent storage adopts a versioned storage strategy, and a new model version is generated each time it is stored, and the differences between versions are recorded.

[0042] The specific training process of the large language model includes: first, pre-training the BERT-based model on a general Chinese corpus; then, collecting ship-related text corpora, including the International Convention for Preventing Collisions at Sea, ship operation manuals, port regulations, and historical ship accident reports, to construct a domain corpus of approximately 5 million words, and using this corpus to continue pre-training the model; finally, using 10,000 manually annotated natural language requirements for ship modeling and their corresponding structured element labels, supervising fine-tuning of the model in downstream intent recognition and element extraction tasks.

[0043] Specifically, the semantic similarity calculation in step S3 includes: For each element in the set of structured modeling elements extracted from user needs and the metamodel description in the metamodel library, the cosine similarity between their word vector representations is calculated. The word vectors are obtained by training the Word2Vec model on a corpus in the shipping domain. Simultaneously, based on the pre-constructed ship domain ontology, the shortest path distance between the element and the corresponding concept in the meta-model description in the ontology graph is calculated and converted into a semantic relevance score, wherein the formula for calculating the semantic relevance score is: ; Where d represents the shortest path distance between the element and the corresponding concept in the metamodel description in the ontology graph. This is the attenuation factor, typically ranging from 1.5 to 2.0, used to control the rate at which the correlation decays due to distance. Finally, the semantic similarity is obtained by weighted fusion of the cosine similarity and the semantic relevance score, and the fusion formula is as follows: ; The weights w1 and w2 are dynamically adjusted according to the modeling context, and w1 + w2 = 1.

[0044] For each element in the structured modeling element set extracted from user needs and its meta-model description in the meta-model library, the cosine similarity between their word vector representations is calculated. The word vectors are obtained by training the Word2Vec model on a corpus in the shipbuilding domain. Word2Vec is an efficient word vector representation method that captures the semantic relationships between words and measures the similarity between words by calculating the cosine similarity between their word vectors.

[0045] Based on a pre-constructed ship domain ontology, the shortest path distance between the element and its corresponding concept in the meta-model description within the ontology graph is calculated and converted into a semantic relevance score. This method measures semantic similarity by calculating the shortest path between two entities in the knowledge graph. Finally, the semantic similarity is obtained by weighted fusion of the cosine similarity and the semantic relevance score. This method combines word vector-based and knowledge graph-based semantic similarity calculation methods, enabling a more comprehensive measurement of semantic similarity.

[0046] In summary, the semantic similarity calculation in step S3 combines word vector representation, ontology knowledge graph, and weighted fusion method to achieve multi-dimensional evaluation of semantic similarity.

[0047] In step S4, the multi-criteria optimization screening is achieved by solving a multi-objective optimization problem. The objective function of this optimization problem aims to maximize the overall fit utility value, which is defined by the following formula: ; Where U represents the overall fit utility value, n is the number of candidate meta-models, and S i C represents the semantic fit between the i-th candidate metamodel and the user's needs. i R represents the coupling degree between the i-th candidate metamodel and other metamodels in the sequence. i denoted as the normalized reuse frequency of the i-th candidate metamodel in historical modeling projects, where α, β, and γ are weight coefficients dynamically determined using the Analytic Hierarchy Process (AHP).

[0048] Traditional meta-model selection often relies on a single metric (such as semantic similarity), neglecting coupling degree and reuse frequency. This formula introduces a multi-objective weighted utility function, which integrates semantic fit, coupling degree, and reuse frequency to improve the overall fit quality of meta-model combinations; avoids the explosion of model complexity caused by high coupling; and encourages the reuse of high-frequency, high-quality meta-models to improve modeling efficiency.

[0049] In this embodiment of the invention, it is a multi-objective optimization problem, the objective of which is to maximize an overall fit utility value (U), which is composed of a weighted combination of multiple factors. Specifically, the formula includes the following elements: U: Overall fit utility value, which is the objective function.

[0050] n: The number of candidate meta-models.

[0051] S i : The semantic fit between the i-th candidate metamodel and the user's needs.

[0052] C i : The coupling degree between the i-th candidate metamodel and other metamodels in the sequence.

[0053] R i : Normalized reuse frequency of the i-th candidate metamodel in historical modeling projects.

[0054] α, β, γ: weighting coefficients, dynamically determined using AHP (Analytic Hierarchy Process).

[0055] Its solution process belongs to multi-objective optimization, and its core lies in selecting the optimal candidate meta-model by optimizing the objective function (U).

[0056] As can be seen, the multi-objective optimization problem is a typical multi-objective optimization problem, the goal of which is to select the optimal candidate meta-model by optimizing the objective function (U).

[0057] The dynamic semantic verification in step S6 includes: The state transition paths defined for behavioral meta-model instances are verified based on linear temporal logic (LTL) formulas, which include safety regulations in the maritime domain, such as collision avoidance rules and channel occupancy agreements. The system automatically traverses the possible behavioral trajectories of meta-model instances using model detection technology to check for violations of preset maritime safety regulations or business rules; for detected conflicts, it generates conflict reports and provides correction suggestions.

[0058] Linear Temporal Logic (LTL) is a type of temporal logic used to describe the dynamic behavior of a system. It represents the state transition path of the system through linear time series. LTL formulas can express the behavior of the system at different points in time, such as the "always" (G), "future" (F), and "until" (U) time series operators, used to describe the safety and liveness properties of the system. LTL is widely used in formal verification and model checking to verify whether a system meets specific specifications.

[0059] Model checking is an automated verification technique that checks whether a system meets predefined properties by comparing its model with specifications (such as LTL formulas). Model checking can automatically traverse all possible behavioral paths of the system to check for violations of specifications. For example, SPIN is a commonly used model checking tool for verifying the correctness of distributed systems.

[0060] In step S6, the core of dynamic semantic verification is to verify system behavior through LTL formulas. For example, LTL formulas can represent safety regulations in the maritime field, such as collision avoidance rules and channel occupancy agreements. Through model detection technology, the system can automatically check whether behavioral metamodel instances violate these rules and generate conflict reports and correction suggestions.

[0061] When a violation of the specification is detected, the system generates a conflict report and provides corrective suggestions. This process relies on the combined use of model checking tools and LTL formulas.

[0062] In summary, the dynamic semantic verification in step S6 is an automated process that combines LTL, model checking, and formal verification to ensure that the system behavior conforms to preset security and business rules.

[0063] During the dynamic semantic verification process, for detected potential conflict behavior trajectories, the system will activate a trajectory correction function. This function generates correction suggestions based on the conflict severity and correction cost. The output of the trajectory correction function is calculated by the following formula: ; Where Δ represents the recommended trajectory correction amount. Represents the set of all possible alternative behavioral trajectories. The original trajectory where a conflict was detected represents the original behavior trajectory, path' represents an alternative behavior trajectory, ω1 and ω2 are weighting factors, the D function is used to quantify the difference in semantic conflict severity between the original trajectory and the alternative trajectory, and the C function is used to calculate the cost required for state transition in a specific state; in the trajectory correction function, D(path) original , path') is the conflict severity difference function, which is the difference in the number of safety rules violated by the alternative trajectory path' and the original conflict trajectory path; C(state, trans(state)) is the state transition cost function, which is quantified according to the computing resources or time consumed by the transition; ω3 and ω4 are weighting factors, set to 0.7 and 0.3 respectively, to balance the conflict resolution and execution costs.

[0064] Traditional consistency checks often rely on manual correction after conflicts are detected, lacking an automated correction mechanism. This formula introduces a multi-objective trajectory correction strategy to balance the cost of conflict resolution with behavioral consistency, achieving automated behavioral trajectory correction; improving the dynamic semantic consistency of the model; and reducing the cost and time of manual correction.

[0065] The process, following step S5 and preceding step S6, includes step S5a: performing a complexity assessment on the initially instantiated metamodel. If the assessment value exceeds a preset threshold, a metamodel simplification operation is triggered. This simplification operation includes merging metamodel instances with similar functions, removing redundant attributes, or abstracting low-level behaviors. The complexity assessment is calculated based on the number of nodes and relations in the metamodel instance, using the following formula: ; in, The number of nodes in the instance. For the number of relations, The total number of possible elements.

[0066] In step S7, the persistent storage adopts a versioned storage strategy. Each storage generates a new model version and records the differences between versions. The configuration management information includes the creator of the model element, the creation time, the last modification time, and the dependency relationship with other model versions. The versioned storage is managed using a Git-like system, which supports version rollback and difference comparison.

[0067] Versioned storage strategies mean that each storage operation generates a new model version and records the differences between versions. This strategy helps track the evolution of the model, facilitating backtracking and management. A versioned storage scheme supporting historical models is mentioned, which states that any backend (such as key-value stores, relational databases, Git, etc.) can be used to store versioned model elements. It also mentions that this scheme supports versioned storage of historical models at runtime.

[0068] The versioned storage is managed using a Git-like system, supporting version rollback and difference comparison. Git is a widely used version control system that supports branch management, version rollback, and difference comparison. This paper introduces the installation and use of the Git version management system, including basic operations, branch management, and remote management. A model version control scheme based on Git principles is proposed to solve the challenges of managing large-scale version systems, supporting version identification, branch management, version difference analysis, rollback, and merging.

[0069] Configuration management information includes the creator, creation time, last modification time, and dependencies of model elements. This information helps track the model's change history and dependencies. The mention of state dependencies in configuration management, such as "who do I depend on" and "who depends on me," is related to dependency management within configuration management.

[0070] Version control tools (such as RCS and Git) are essential for implementing version control. RCS (Revision Control System) is mentioned as a software tool for version control, supporting version management of text files. Source Code Control Systems are also mentioned, being one of the earlier version control systems. The implementation and evaluation of a Revision Control System are discussed.

[0071] Version control supports not only individual development but also team collaboration. It is mentioned that version control is used in team environments to coordinate development models, supporting version storage, comparison, and rollback. The proposed model version control scheme supports distributed collaboration and automated testing pipelines to improve team collaboration efficiency.

[0072] In summary, the persistent storage strategy and version management mechanism described in step S7 combine version control, Git-like system, configuration management, version rollback and difference comparison functions, aiming to achieve efficient management and collaboration of model versions.

[0073] The construction of the ship domain system meta-model library in step S1 also includes: dynamically updating the meta-model based on historical AIS data, wherein AIS data is used to extract ship behavior patterns and optimize the behavior meta-model rules through machine learning algorithms (such as cluster analysis).

[0074] AIS (Automatic Identification System) data is a core data source for ship behavior analysis. Preprocessing (such as data cleaning, resampling, and normalization) can improve data quality and provide a reliable foundation for subsequent analysis. For example, literature mentions preprocessing raw AIS data, including filtering, interpolation, and compression, to generate a ship behavior dataset that can be used for analysis.

[0075] Cluster analysis is a key method for identifying ship behavior patterns. By clustering historical AIS data using clustering algorithms (such as K-means, UMAP, and spectral clustering), different types of ship behavior patterns can be identified. UMAP and spectral clustering algorithms are used to reduce the dimensionality and cluster ship behavior to identify different behavior patterns. The clustering results can be used to build a behavior pattern library, providing a foundation for subsequent model optimization.

[0076] Based on clustering results, behavioral meta-models can be constructed and optimized. For example, the literature mentions using sequence-to-sequence models and attention mechanisms to predict ship behavior. Furthermore, machine learning algorithms (such as the EM algorithm) can be used to continuously optimize model parameters to adapt to new data features, thereby continuously improving anomaly detection and behavior prediction capabilities.

[0077] The construction of the system metamodel library is a dynamic process. By continuously introducing new AIS data, the system can continuously update behavioral patterns and model rules. When new anomalous behaviors are detected, the system updates the model to adapt to the new data characteristics, thereby continuously optimizing anomaly detection capabilities. This dynamic update mechanism ensures that the model can adapt to constantly changing ship behavior patterns.

[0078] The specific training process of the large language model in step S2 includes: pre-training using ship domain text corpus and fine-tuning it to adapt to the natural language recognition requirements; the domain text corpus includes ship manuals, navigation rules and accident reports.

[0079] The training process of a large language model is usually divided into two main stages: pre-training and fine-tuning.

[0080] The pre-training phase is the first step in training a large language model. Its purpose is to train the model on large-scale unlabeled text data to learn general representations and linguistic patterns. This phase typically uses self-supervised learning methods, such as next-word prediction or masked language models. The goal of pre-training is to enable the model to learn the statistical regularities, syntactic structures, and semantic relationships of the language so that it can be applied to downstream tasks.

[0081] During the pre-training phase, the model extracts text segments from a large-scale text dataset and trains it using the stochastic gradient descent (SGD) algorithm. The pre-training text data can be a general corpus (such as a general English corpus) or domain-specific text, such as ship manuals, navigation rules, and accident reports.

[0082] The fine-tuning phase involves further training the pre-trained model using a task-specific dataset to adapt it to the specific requirements of that task. The fine-tuning phase typically employs supervised learning methods, adjusting the model with a small amount of labeled data to improve its performance on the specific task.

[0083] During the fine-tuning phase, the model is adjusted according to the needs of specific tasks, such as natural language processing, text classification, and question answering. The fine-tuning phase typically has low computational costs because the amount of data is small and the model already possesses general language representation capabilities.

[0084] The "ship domain text corpus" mentioned in this invention includes ship manuals, navigation rules, and accident reports. These texts can serve as input data for either the pre-training or fine-tuning phases. During pre-training, these domain texts can be used as part of the pre-training data to help the model learn ship-related language and knowledge. During fine-tuning, these domain texts can be used to fine-tune the model to improve its performance on ship-related tasks such as natural language requirement recognition.

[0085] Therefore, the training process of a large language model includes two stages: pre-training and fine-tuning. The pre-training stage is conducted on large-scale unlabeled text to learn general language representations; the fine-tuning stage is conducted on task-specific data to adapt to the needs of specific tasks. The "ship domain text corpus" mentioned in this invention can be used as input data for either the pre-training or fine-tuning stage to improve the model's performance on ship-related tasks.

[0086] This invention also proposes an intelligent meta-model construction and dynamic adaptation system for the shipbuilding field using the above-mentioned method, characterized by comprising: The metamodel library management module is used to store and manage the metamodel library of the ship domain system, and provides functions for adding, deleting, modifying and querying metamodels; The natural language processing module integrates a large language model and is used to perform semantic understanding and modeling intent recognition in step S2. The meta-model matching and recommendation module is used to perform semantic similarity calculation and meta-model matching in step S3; The optimization filtering module is used to perform multi-criteria optimization filtering in step S4. The instantiation module is used to perform the metamodel instantiation and association mapping in step S5; The consistency verification module is used to perform the consistency verification in step S6; The storage management module is used to perform persistent storage and version management in step S7.

[0087] The metamodel library management module is used to store and manage the metamodel library of shipbuilding systems, providing functions for adding, deleting, modifying, and querying metamodels. Based on [specific criteria], the metamodel management module is used to operate on metamodels. A metamodel is a definition of various types of metadata and the relationships between metadata, including standard metamodels and user-defined metamodels. The metamodel management module supports adding, deleting, modifying, querying, and version management of metamodels, consistent with the functional description of the metamodel library management module.

[0088] The Natural Language Processing (NLP) module integrates a large language model for performing semantic understanding and modeling intent recognition. Although the evidence does not directly mention the NLP module, large language models have wide applications in semantic understanding and intent recognition, therefore the functional description of this module is reasonable.

[0089] The metamodel matching and recommendation module is used to perform semantic similarity calculation and metamodel matching. According to [reference needed], metamodel management involves the unified management of metamodels, lifecycle management, and the storage and management of modeling artifacts. Metamodel matching and recommendation functions can be implemented based on metamodel similarity.

[0090] The optimization filtering module is used for multi-criteria optimization filtering. Although the evidence does not directly mention the optimization filtering module, the optimization and filtering logic involved in functions such as metamodel management and metadata management can support the implementation of this module.

[0091] The instantiation module is used for metamodel instantiation and association mapping. According to [the relevant documentation], the model library system supports model storage, operation, and maintenance, including model instantiation and association mapping, which aligns with the functional description of the instantiation module.

[0092] The consistency verification module is used for consistency verification. According to [the relevant documentation], metadata management includes functions such as metadata change management, version management, and quality verification; the consistency verification module can be implemented based on these functions.

[0093] The storage management module is used for persistent storage and version management. According to [reference needed], the metadata management platform supports metadata storage, version management, backup and recovery functions, which aligns with the functional description of the storage management module.

[0094] The system also integrates a ship AIS data interface for receiving AIS data in real time and updating the meta-model library.

[0095] An AIS (Airborne Information System) is a system that automatically sends and receives data via shipboard equipment for communication and information exchange between vessels. AIS equipment acquires vessel location information through GPS and transmits and receives data via radio waves, thereby enabling real-time communication and information sharing between ships. AIS systems are not only used for communication between ships but are also widely applied in areas such as ship traffic management, navigation safety, and port management.

[0096] The reception and processing of AIS data typically involves multiple technical stages. For example, AIS data can be received via land-based receivers, satellite receivers, or shipboard equipment, and exchanged with computer systems through interfaces. AIS data processing generally includes steps such as data decoding, data storage, and data visualization. For instance, AIS data can be converted into location information in a Geographic Information System (GIS) for real-time monitoring and analysis of ship movements.

[0097] The real-time nature of AIS data is one of its key characteristics. The update frequency of AIS data depends on the ship's speed, position, and communication conditions, typically updating every 2-10 seconds. Although AIS data is real-time, its updates are not strictly time-driven but event-driven, meaning they are updated based on the ship's activities and communication conditions.

[0098] The accuracy of AIS data also has certain limitations. AIS data may contain errors or be delayed, such as errors in data transmission or delays in data processing. Therefore, in practical applications, AIS data usually needs to be used in conjunction with other data sources to improve data accuracy and reliability.

[0099] The integrated AIS data interface of the system can receive AIS data and update the meta-model library in real time, which relies on the data transmission, processing, and storage capabilities of the AIS system. Through its wide application and real-time performance, the AIS system provides crucial data support for ships and related systems.

[0100] The consistency verification module also includes a visualization tool for graphically displaying verification results and conflict trajectories.

[0101] Example 3 This disclosure provides a non-volatile computer storage medium storing computer-executable instructions that can perform the steps described in the above embodiments.

[0102] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0103] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0104] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (AN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0105] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0106] The units described in the embodiments of this disclosure can be implemented in software or hardware. The names of the units are not, in some cases, intended to limit the specific unit.

[0107] The preferred embodiments of the present invention have been described above to make the spirit of the present invention clearer and easier to understand, and are not intended to limit the present invention. All modifications, substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope summarized by the appended claims.

Claims

1. A method for constructing and dynamically adapting an intelligent meta-model for the shipbuilding field, characterized in that, Includes the following steps: S1. Construct a meta-model library for shipbuilding systems, which includes conceptual meta-models, interaction meta-models, and behavioral meta-models. S2. Utilize a large language model to perform semantic understanding and modeling intent recognition on the natural language input from users, extract modeling elements, and form a structured set of modeling elements. S3. Based on the set of structured modeling elements obtained in step S2, perform meta-model matching in the meta-model library through semantic similarity calculation to generate a set of candidate meta-models; S4. Perform multi-criteria optimization and screening on the candidate meta-model set generated in step S3, and output the optimal meta-model adaptation sequence. S5. Based on the optimal metamodel adaptation sequence obtained in step S4, instantiate metamodel instances and construct the association mapping relationship between metamodel instances. S6. Perform a consistency check on the set of meta-model instances generated in step S5; S7. Persistently store the verified meta-model instances and their relationships in the ship domain model library, and generate corresponding model documents and configuration management information.

2. The method as described in claim 1, characterized in that, In step S1, the conceptual meta-model is used to define entities and their static attributes within the ship domain, the interaction meta-model is used to define dynamic interaction relationships and protocols between entities, and the behavioral meta-model is used to define the state transition logic and behavioral rules of entities; and the construction of the meta-model library is based on historical AIS data, ship navigation logs, and domain ontology.

3. The method as described in claim 2, characterized in that, In step S1, the conceptual meta-model includes ship, port, and waterway entities, the interaction meta-model includes ship-port communication protocols, and the behavioral meta-model includes ship navigation state change rules.

4. The method as described in claim 1, characterized in that, In step S2, the large language model is trained based on the BERT architecture and integrates a marine domain dictionary; the extracted modeling elements include domain entity keywords, interaction relationship descriptions, and behavioral constraints.

5. The method as described in claim 1, characterized in that, In step S3, the semantic similarity calculation integrates word vector cosine similarity with semantic distance based on the ship domain ontology.

6. The method as described in claim 1, characterized in that, In step S4, the multi-objective decision model achieves optimized screening by maximizing the overall fit utility value. The overall fit utility value is a weighted function of semantic fit, coupling degree, and normalized reuse frequency, wherein the weight coefficients are dynamically determined by the AHP (Analytic Hierarchy Process).

7. The method as described in claim 1, characterized in that, After step S5 and before step S6, step S5a is also included: performing a complexity assessment on the initially instantiated metamodel, the complexity being calculated based on the number of nodes and relations of the metamodel instance; if the assessment value exceeds a preset threshold, a metamodel simplification operation is triggered, the simplification operation including merging metamodel instances with similar functions, removing redundant attributes, or abstracting low-level behaviors.

8. The method as described in claim 1, characterized in that, In step S6, the consistency check includes static syntax check, dynamic semantic check, and cross-view logic consistency check. The dynamic semantic verification includes: verifying the state transition path based on linear temporal logic formulas, and automatically traversing the behavior trajectory using model detection technology to check whether there is a violation of safety regulations in the maritime field.

9. The method as described in claim 1, characterized in that, In step S7, the persistent storage adopts a versioned storage strategy, generating a new model version each time it is stored, and recording the differences between versions; the configuration management information includes the creator of the model element, the creation time, the last modification time, and the dependencies.

10. A ship-oriented intelligent meta-model construction and dynamic adaptation system that implements the methods of claims 1-9 above, comprising: The meta-model library management module is used to store and manage the meta-model library of the ship domain system. The natural language processing module is used to perform semantic understanding and modeling intent recognition in step S2; The meta-model matching and recommendation module is used to perform semantic similarity calculation and meta-model matching in step S3; The optimization filtering module is used to perform multi-criteria optimization filtering in step S4. The instantiation module is used to perform the metamodel instantiation and association mapping in step S5; The consistency verification module is used to perform the consistency verification in step S6; The storage management module is used to perform persistent storage and version management in step S7.

Images