Complex equipment system architecture retrieval and adaptation method based on matter-element theory and CBR
By using a system architecture retrieval and adaptation method based on matter-element theory and case-based reasoning, the problem of insufficient model knowledge reuse in MBSE is solved, enabling rapid generation and reliable reuse of complex equipment system architectures, thereby improving design efficiency and automation levels.
Patent Information
- Application Number
- CN202511267862.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-05
- Publication Date
- 2026-01-06
AI Technical Summary
The existing MBSE process suffers from insufficient model knowledge reuse and excessive reliance on human experience, resulting in low design efficiency and serious problems of repeatedly building similar models, making it difficult to achieve rapid generation and reliable reuse of complex equipment system architectures.
A system architecture retrieval and adaptation method based on matter-element theory and case reasoning is adopted. Through SysML modeling and matter-element description, combined with grey relational analysis and information entropy theory, a similarity calculation model is constructed to achieve structured representation and intelligent adaptation of existing system architecture models.
It enables effective structured storage and efficient retrieval of existing system architecture models, significantly improving the generation efficiency of complex equipment system architectures, reducing development costs and expert dependence, and enhancing the reliability and automation level of the design.
Smart Images
Figure CN121277931A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the technical field of system architecture design, and particularly relates to a complex equipment system architecture retrieval and adaptation method based on the matter element theory and CBR. BACKGROUND
[0002] With the deepening application of emerging technologies in the field of equipment, the functional requirements of modern equipment systems continue to improve, and the complexity of systems also grows exponentially. The traditional document-based system design method has been difficult to meet the requirements of high complexity and multi-disciplinary coupling of equipment development due to long development cycle, low collaboration efficiency, and weak information integration capability. Under this background, model-based systems engineering (MBSE) has gradually become the mainstream paradigm for the development of complex systems. MBSE takes system models as the core, supports the multi-dimensional expression and optimization of system functions, architectures, and behaviors through continuous verification, iteration, and simulation, and has been widely applied in safety-critical fields such as aviation, aerospace, and nuclear energy.
[0003] However, the current MBSE practice still highly depends on the experience and knowledge of domain experts to complete the construction, verification, and integration of system models, and has not yet realized the effective extraction and reuse of design knowledge contained in existing system models. On the one hand, existing successful models are stored as isolated instances, and the correlation between their requirement mapping, functional structure, and performance parameters cannot be structured and reused. On the other hand, when facing new design tasks, designers often repeatedly construct models with similar functions or architectures, and the phenomenon of "repeating the same mistakes" is significant, resulting in waste of knowledge resources and reduction of development efficiency. In addition, the model adaptation process based on artificial experience is highly subjective and inconsistent, which also limits the reliability and reusability of system architecture design.
[0004] The matter element theory provides a formalized description method for the attributes, numerical values, and relationships of physical objects, supports multi-dimensional modeling and extension analysis of complex objects, and has been successfully applied in fields such as fault diagnosis and scheme evaluation. Case-based reasoning (CBR) is a problem-solving method based on historical experience, which can provide reusable solutions for new problems by retrieving and adapting similar cases, and is particularly suitable for engineering fields with high experience dependence and high model reuse demand.
[0005] Therefore, in view of the problems of insufficient model knowledge reuse, high artificial dependence, and low architecture construction efficiency in the existing MBSE process, it is urgent to build a systematic model retrieval and reuse mechanism that can effectively represent, efficiently retrieve, and intelligently adapt existing system architecture models, thereby supporting the rapid generation of complex equipment system architectures and the continuous accumulation of knowledge. SUMMARY
[0006] In view of the problems of insufficient reuse of existing system architecture models, excessive dependence on manual experience and low design efficiency in the existing MBSE process, the application proposes a system architecture retrieval and adaptation method based on the matter element theory and case-based reasoning. The application aims to realize effective structured representation and storage of system model knowledge, and through intelligent retrieval and adaptation mechanism, to support rapid generation and reliable reuse of complex equipment system architecture, thereby improving design efficiency and reducing development cost and dependence on human subjective experience.
[0007] In an aspect of the application, a system architecture retrieval and adaptation method based on the matter element theory and case-based reasoning is provided, comprising the following steps: S1, modeling the equipment system architecture based on SysML, and structuring the model information based on the matter element theory to build an equipment system architecture model instance library; S2, for a new design requirement, describing its function and performance requirements in the form of matter elements as a target instance, constructing a similarity calculation model of the target instance based on gray correlation analysis and information entropy theory, retrieving similar cases from the instance library and adapting them, and outputting a feasible system architecture scheme.
[0008] In one embodiment, step S1 comprises the following steps: S11, modeling the existing equipment system based on SysML, SysML describes the system architecture scheme of the equipment through module graph, internal module graph and attribute value of module; S12, based on the matter element theory, describing the SysML model in the form of matter elements, defining the attribute set of the module object N , A =( I , R , P , H ); wherein, I is N a unique number; R is N a satisfaction relationship to the requirement; P is N an inherent attribute of the module, including a structure attribute and an operation attribute; H is N hierarchical information of the module, indicating the hierarchical relationship in the system architecture.
[0009] In one embodiment, step S2 comprises the following steps: S21, constructing a case retrieval model, and representing the feature vectors of the target instance and the alternative instance as v r N* and v rNi A relationship matrix between the two is established, and a similarity calculation method combining Hamming distance and Mahalanobis distance is adopted. S22. Calculate feature weights based on entropy weight theory, construct a decision matrix and perform dimensionless processing, and calculate the evaluation entropy and entropy weight of each indicator. S23. Calculate the correlation coefficient between the target problem and the candidate instances based on grey relational analysis, and determine the instance model with the highest similarity. S24. Based on the SysML model objects obtained from the retrieval, perform model adaptation according to the adaptation rules, including generating sub-models, associating models, modifying attribute values, and integrating with the system model.
[0010] In one implementation, the similarity calculation formula is:
[0011] in, S It is the first of the alternative instances i The covariance matrix of the eigenvector groups; D ij The Mahalanobis distance between two numerical vectors; P ij The similarity between numerical vectors; AND operation; x i These are the eigenvectors that have undergone dimensionless processing; s ij The overall similarity between the mixed vectors; w 1 and w 2 represents the weights of the numerical part and the encoded part of the vector, respectively; f i For the target instance i One functional requirement; c i For the target instance i One performance requirement; f i j For the first j The first of the alternative instances i One functional requirement; c j i For the first j The first of the alternative instances i Performance requirements.
[0012] In one implementation, the dimensionless processing of the decision matrix is performed using the formula:
[0013] in, x ijFor the first j The first of the alternative instances i Attribute values ( i When the sum is 1, 2, ... m, it represents the functional attribute value. i When the value is (m+1), (m+2), ... (m+n), it represents the performance attribute value. x max i For the decision matrix of the th i The maximum value in the column; x min i For the decision matrix of the th i The minimum value in the column; r ij It is standardized x The corresponding value; c is a constant to prevent null values.
[0014] In one implementation method, the formula for calculating the correlation coefficient is:
[0015] in, ρ The resolution coefficient; r 0j The first reference vector of the target problem j One element, r ij The first vector represents the candidate instance comparison vector. j Each element.
[0016] In one implementation, the adaptation rules in step S24 include: Sub-models based on component attribute vector generation and retrieval instances with directed combination relationships; A model that generates and retrieves instances with related relationships based on reference attribute vectors; Modify the value attribute of the retrieved instance based on the demand vector of the target instance; The retrieved instances are integrated into the system model based on hierarchical relationships.
[0017] In another aspect of this application, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the above-described system architecture retrieval and adaptation method based on matter-element theory and case reasoning.
[0018] In another aspect of this application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the above-described system architecture retrieval and adaptation method based on matter-element theory and case reasoning.
[0019] The beneficial effects of this application are as follows: 1) It realizes the structured extraction and storage of design knowledge from existing system architecture models. By using matter-element theory to uniformly represent the attributes, relationships and levels of SysML models, a reusable instance library is constructed, overcoming the problems of scattered model knowledge and difficulty in direct reuse in traditional MBSE.
[0020] 2) A hybrid similarity metric method integrating Hamming distance and Mahalanobis distance is proposed, which can simultaneously handle mixed input conditions of functional coding and performance values, and significantly improves the adaptability and accuracy of case retrieval.
[0021] 3) Combining grey relational analysis and information entropy weight theory, a more reliable case similarity evaluation model was established, which fully considers the correlation between attributes and the distribution of information value, making the retrieval results more meaningful for engineering practice.
[0022] 4) It provides a systematic case adaptation mechanism, which supports automatic adjustment and integration of models based on components, references, attributes and hierarchical relationships, significantly reducing manual intervention and improving the automation level and efficiency of complex system architecture construction.
[0023] 5) A complete system architecture reuse framework based on case-based reasoning has been formed, which can significantly shorten the equipment system design cycle, reduce development costs and expert dependence, and promote the transformation of MBSE from human experience-driven to intelligent reuse. Attached Figure Description
[0024] Figure 1 This is a flowchart illustrating the system architecture retrieval and adaptation method based on matter-element theory and case reasoning in this application. Detailed Implementation
[0025] The technical solution of this application will be clearly and completely described below with reference to specific embodiments. However, those skilled in the art will understand that the embodiments described below are only some embodiments of this application, not all embodiments, and are only used to illustrate this application, and should not be regarded as limiting the scope of this application. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0026] In one embodiment, a system architecture retrieval and adaptation method based on matter-element theory and case reasoning is provided, referring to... Figure 1 As shown, the process includes: modeling the equipment system architecture based on SysML, and storing the model information in a structured manner according to the matter-element theory to build an equipment system architecture model instance library; for new design requirements, describing their functional and performance requirements in the form of matter-element as target instances, constructing a similarity calculation model for the target instances based on grey relational analysis and information entropy theory, retrieving similar cases and adapting them, and outputting feasible system architecture solutions.
[0027] Specifically, a matter-element model is used to structurally represent the attributes, operations, and relationships of SysML modules. This matter-element model includes the following elements: unique identifiers, requirement fulfillment relationships, structural and operational attributes, and hierarchical relationships. Through these elements, the various modules in the system architecture and their interrelationships can be clearly and structurally expressed and stored.
[0028] In some embodiments, the specific implementation process includes the following steps: 1) Model the existing equipment system using SysML, and construct the system architecture scheme of a type of equipment by using Block Definition Diagram (BDD), Internal Block Diagram (IBD) and attribute value description.
[0029] In this scheme, a certain function or performance of the system is represented by a SysML module, which is mapped to the corresponding subsystem or functional component in the physical structure of the system.
[0030] Each module in a SysML model has the following two types of attributes: Structural attributes include constraints, flows, parts, references, and values, which describe the static characteristics and compositional relationships of modules. Operational attributes: These include operations and receptions, and are used to describe the behavior and interaction capabilities of a module.
[0031] These attributes and their values are a response to design requirements, and also describe the position and relationship of the module in the system structure, providing a basis for subsequent system verification and validation.
[0032] 2) Based on the matter-element theory, each module in the SysML model above is described using matter-element concepts. For any given module object... N Define its attribute set as A =( I , R , P , H ); in, I Representation module N A unique identifier, used to uniquely identify the module in the instance library; R Representation module NThe relationship between the design requirements and the system requirements is defined by the value of "satisfy" or "achieve". P Representation module N The inherent properties of SysML models include the aforementioned structural and operational properties, which are directly derived from the definitions in SysML models. H Representation module N The hierarchical information describes the module's hierarchical position within the overall system architecture and its compositional relationship with other modules. This information is related to... P The component attributes are associated with it.
[0033] In this implementation, a hybrid strategy of Hamming distance and Mahalanobis distance is employed for similarity calculation to address mixed input conditions that simultaneously include numerical performance requirements and coding functional requirements. In the similarity calculation model, feature weights are determined based on information entropy theory, and grey relational analysis is used to evaluate the degree of association between the target instance and candidate instances, thereby improving the reliability of the retrieval results.
[0034] In this embodiment, case adaptation includes at least one of the following operations: generating sub-models based on component attributes, generating association models based on reference attributes, modifying attribute values based on target requirements, and integrating into the system architecture based on hierarchical relationships.
[0035] In some embodiments, the specific steps include: 1) Constructing a case retrieval model and calculating similarity Design Goal Example N * The demand feature vector is: v r N* =( f 1, f 2 ... f m ; c 1, c 2 ... c n ) in, f i The first instance represents the target instance. i Each functional requirement takes a Boolean value (encoded as 0 or 1). c i The first instance represents the target instance. i Each performance requirement is a numerical indicator. m and n These represent the number of functional and performance requirements, respectively.
[0036] Correspondingly, alternative instancesN j The feature vectors are: v r Nj =( f 1 j , f 2 j ... f m j ; c 1 j , c 2 j ... c n j ) in, f i j Indicates alternative instances N j The i Each functional requirement takes a Boolean value (encoded as 0 or 1). c i j Indicates alternative instances N j The i Each performance requirement is a numerical indicator.
[0037] Case similarity converted to the above v r N* and v r Nj A similarity measure between two vectors. Therefore, establish... v r N* and v r Nj Relationship matrix R :
[0038] in, s i ( i =1,2 … m ) is the first of the target cases i Functional requirements f i and the j The first of the alternative instances i The similarity between functional requirements; s i ( i =( m +1), (m +2)... ( m + n ()) is the first of the target cases i Performance requirements f i and the j The first of the alternative instances i The similarity between performance requirements.
[0039] To handle mixed coded and numerical data and ensure the similarity between each element and the correlation between variables, a similarity calculation method combining Hamming distance and Mahalanobis distance is proposed, as follows: (1) in, S It is the first of the alternative instances i The covariance matrix of the eigenvector groups; D ij The Mahalanobis distance between two numerical vectors; P ij The similarity between numerical vectors; AND operation; xi These are the eigenvectors that have undergone dimensionless processing; s ij The overall similarity between the mixed vectors; w 1 and w 2 represents the weights of the numerical part and the encoded part of the vector, respectively; f i For the target instance i One functional requirement; c i For the target instance i One performance requirement; f i j For the first j The first of the alternative instances i One functional requirement; c j i For the first j The first of the alternative instances i Performance requirements. It is evident that the smaller the distance between vectors, the greater their similarity.
[0040] 2) Determining feature weights based on information entropy theory Set target instance N* have m+n Features, and for instance problems N* have k Given 10 alternative instances, construct the decision matrix X = ( x ij )(m+n)×k , ( i =1, 2, ... (m+n), j =1, 2, ... k X is dimensionless, specifically by: (2) in, x ij For the first j The first of the alternative instances i Attribute values ( i When the sum is 1, 2, ... m, it represents the functional attribute value. i When the value is (m+1), (m+2), ... (m+n), it represents the performance attribute value. x max i For the X matrix, the first... i The maximum value in the column; x min i For the X matrix, the first... i The minimum value in the column; r ij It is standardized x The corresponding value; c is a fixed constant to prevent null values from appearing during the process of calculating entropy weights.
[0041] No. i The entropy value of an index is defined as: (3) in, t ij It is the first i One alternative instance j The attribute value is the [number]th among all candidate instances. j The proportion of the sum of attribute values is used to characterize the magnitude of the variation in that attribute value; K The coefficient for calculating information entropy when there are (m+n) attributes; H j The first in the alternative instance library j Information entropy of each attribute.
[0042] Then the first j The evaluation entropy weight of each indicator is: (4) in, m The number of functional attributes for alternative instances; n The number of performance attributes for candidate instances; H j The first in the alternative instance library j The information entropy value of each attribute; wj For the first j The entropy weight of each attribute.
[0043] 3) Case retrieval based on grey relational analysis: The degree of correlation between the target problem and candidate instances is calculated using grey relational analysis. Let the feature vector of the target instance be the reference sequence. R 0, the feature vector of the candidate instance is the comparison sequence. R 1. Calculation R 0( j )and R i ( j The correlation coefficient between §( j ): (5) Where 0≤ρ≤1 is the resolution coefficient, and in general, ρ=0.5 is taken; r 0j It is a reference vector R 0th j One element, r ij Represents comparison vectors R i The j Each element.
[0044] The correlation degree of the weighted combination is calculated as shown in equation (6): (6) in, a i Indicates the first i The similarity between candidate instances and the target instance is used for case ranking and retrieval; §( j ) is the target instance and the first i The first of the alternative instances j The degree of correlation between attributes; w j Let be the entropy weight of the j-th attribute.
[0045] 4) Case adaptation based on relation sets For the retrieved SysML model matter elements, according to their matter element characteristics v r The values are used as instance features to calculate similarity with the target problem, resulting in instance models with high similarity. N p Then follow these rules to... N p Adaptation is performed. Adaptation rules include: Based on component attribute vector generation and N pSub-models with directional combination relationships; Based on the generation of reference attribute vectors and N p A model with related relationships; Modify the demand vector based on the target instance. N p The corresponding value attribute; Based on hierarchical relationships, N p It is integrated into the system model.
[0046] Although the embodiments of this application have been described above in conjunction with the accompanying drawings, this application is not limited to the specific embodiments and application fields described above. The specific embodiments described above are merely illustrative and instructive, not restrictive. Those skilled in the art can make many other forms based on the guidance of this specification and without departing from the scope of protection of the claims of this application, and these are all within the scope of protection of this application.
Claims
1. A system architecture retrieval and adaptation method based on the matter element theory and case-based reasoning, characterized in that, The method comprises the following steps: S1, modeling the equipment system architecture based on SysML, and structuring the model information according to the matter element theory, and constructing an equipment system architecture model instance library; S2, for a new design requirement, describing the function and performance requirements as a target instance in the form of a matter element, constructing a similarity calculation model of the target instance based on gray correlation analysis and information entropy theory, retrieving similar cases from the instance library and adapting them, and outputting a feasible system architecture scheme.
2. The system architecture retrieval and adaptation method of claim 1, wherein, Step S1 comprises the following steps: S11, modeling the existing equipment system based on SysML, SysML describes the system architecture scheme of the equipment through module diagram, internal module diagram and attribute value of the module; S12, based on the matter element theory, the SysML model is described in matter element, for module object N , define its attribute set A = ( I , R , P , H ) ; wherein, I is N Unique number; R is N The satisfaction relation of demand; P is N Inherent attribute, including structure attribute and operation attribute; H is N The hierarchical information of, indicates its hierarchical relationship in system architecture.
3. The system architecture retrieval and adaptation method of claim 1, wherein, Step S2 comprises the following steps: S21, constructing a case retrieval model, the feature vectors of the target instance and the alternative instance are represented as v r N* and v r Ni , a relationship matrix is established between the two, and a similarity calculation method of Hamming distance and Mahalanobis distance is adopted; S22, calculating the feature weight based on the entropy weight theory, constructing the decision matrix and performing dimensionless processing, and calculating the evaluation entropy and entropy weight of each index; S23, calculating the correlation coefficient between the target problem and the candidate instance based on gray correlation analysis, and determining the instance model with the highest similarity; S24, based on the SysML model matter element obtained by retrieval, model adaptation is performed according to the adaptation rules, including generating a submodel, associating a model, modifying an attribute value and accessing a system model.
4. The system architecture retrieval and adaptation method of claim 3, wherein, The similarity calculation formula is: wherein, S is the covariance matrix of the i th feature vector group of the alternative instance; D ij is the Mahalanobis distance between two numerical vectors; P ij is the similarity between numerical vectors; is the operation of AND; x i is the dimensionless processed feature vector; s ij is the comprehensive similarity between hybrid vectors; w 1 and w 2 are the weights of the numerical part and the encoding part of the vector, respectively; f i is the i th functional requirement of the target instance; c i is the i th performance requirement of the target instance; f i j is the j th functional requirement of the i th alternative instance; c j i is the j th performance requirement of the i th alternative instance.
5. The system architecture retrieval and adaptation method of claim 3, wherein, The dimensionless processing of the decision matrix adopts the formula: in, x ij For the first j The first of the alternative instances i Each attribute value; x max i For the decision matrix of the th i The maximum value in the column; x min i For the decision matrix of the th i The minimum value in the column; r ij It is standardized x The corresponding value; c is a constant to prevent null values.
6. The system architecture retrieval and adaptation method of claim 3, wherein, The correlation coefficient calculation formula is: in, The adaptation rules include: The resolution coefficient; r 0j The first reference vector of the target problem j One element, r ij The first vector representing the candidate instance comparison vector. j Each element.
7. The system architecture retrieval and adaptation method of claim 3, wherein, Generating a submodel with a directional combination relationship with the retrieved instance based on the component attribute vector; Generating a model with an association relationship with the retrieved instance based on the reference attribute vector; Modifying the value attribute of the retrieved instance based on the requirement vector of the target instance; Accessing the retrieved instance into the system model based on the hierarchical relationship. A computer program is stored thereon, which is executed by the processor to realize the system architecture retrieval and adaptation method of any one of claims 1-7.
8. A computer-readable storage medium, characterized in that, A computer program is stored thereon, which is executed by the processor to realize the system architecture retrieval and adaptation method of any one of claims 1-7.
9. An electronic device, comprising: