Knowledge graph and digital twinning fused electromechanical product reverse design method and system

Abstract

The invention belongs to the technical field of intelligent manufacturing, and particularly relates to a knowledge graph and digital twinning fused electromechanical product reverse design method and system, and a knowledge graph and digital twinning fused electromechanical product reverse design method and system. Semantic unification and dynamic mapping of full-life-cycle multi-source heterogeneous data are achieved, and the reliability and adaptability of the model in the mechanical and electrical product design, manufacturing and operation and maintenance process are enhanced.

Description

Technical Field

[0001] The present application belongs to the field of intelligent manufacturing technology, and specifically relates to a reverse design method and system for electromechanical products that integrates knowledge graphs and digital twins. Background Art

[0002] With the improvement of the automation level of industrial products, users' demand for personalized design and rapid iteration is growing. In the design and development process of complex electromechanical products, a one-way data flow is usually adopted, that is, the product goes through the entire life cycle of design, manufacturing, operation and maintenance until scrapping. In this process, the product life cycle data and experience are not fully utilized, which leads to low design efficiency and reduced product adaptability. Against this background, the reverse design method came into being. The concept of reverse design is to integrate the information data design end at different stages, so that the adaptability of the product is considered at the beginning of the design, so as to achieve product feedback optimization and iterative upgrades.

[0003] Since electromechanical products generate multi-dimensional, structured, and non-uniform data during the design, production, testing, and operation processes, the sharing, flow, and efficient use of knowledge models are hindered. In addition, existing models lack effective data management and semantic association methods, which affects the adaptability and comprehensiveness of design results. In order to solve this problem, the knowledge graph (KG) provides a knowledge management method that can transform, classify, store, and query semantic knowledge at different stages of the product life cycle, effectively solving the problem of information islands. On the other hand, the digital twin (DT) builds a virtual model that is updated synchronously with the physical product, making the flow of data and the evolution of the model possible during the design process. DT can use real-time operating data collected by sensors to dynamically modify the virtual model, thereby accurately reflecting the state of the physical system. Therefore, the present invention combines the structured knowledge management of KG with the dynamic update capability of DT, and proposes a reverse design method and system for electromechanical products. Summary of the invention

[0004] In view of the problems faced in the traditional design process, such as inconsistent formats and semantics of multi-source data, difficulty in dynamically adjusting model parameters, and difficulty in mapping the relationship between models and data, a reverse design method and system for electromechanical products integrating knowledge graph and digital twin is proposed. The technical solution is: A reverse design method for electromechanical products integrating knowledge graph and digital twin includes the following steps: S1. Construct a digital twin model with multi-dimensional information association; The digital twin model is divided into three dimensions: information model, mechanism model and domain model. The metamodels corresponding to the three together constitute a complete digital twin model DTM; S2. Establish a digital twin model information management mechanism based on knowledge graph; In the digital twin model, the interaction between the information model IM, the mechanism model PM and the domain model FM determines the mapping method of the information flow; the knowledge graph helps to achieve information exchange in the product design, manufacturing and operation process by accessing, integrating and converting multi-dimensional information; S3. Establish an information-model mapping mechanism based on knowledge graph; Analyze multi-dimensional information data at different stages of the product life cycle, use knowledge graphs to uniformly manage multi-source heterogeneous information, and establish multiple semantic mapping relationships between information and models; S4. Reverse design optimization process integrating knowledge graph and digital twin; The knowledge graph is used for querying and reasoning about knowledge. It updates the digital twin model by mining hidden design knowledge and demand information through data association, forming a continuous feedback knowledge update loop and enhanced reverse optimization design.

[0005] Preferably, in step S1, the digital twin model expression is as follows: ; Information model IM refers to a logical data set formed by organizing and structuring the original data according to a certain information modeling framework and semantic rules; among them: IMI is the input of digital twin model simulation and mapping, IMO represents the virtual simulation prediction results and system response information obtained in the digital space, and IMDT is the information data generated by the digital twin during the mapping and feedback process; The mechanism model PM is a quantitative description of the physical behavior and internal mechanism of a product or system; The domain model FM is a collection of models built based on the different behavioral characteristics and engineering application requirements of equipment throughout its life cycle.

[0006] Preferably, the digital model information management mechanism based on the knowledge graph established in step S2 includes the following three modules: Data processing module: responsible for processing physical space data in different formats; The key technology module fills the instance information into the corresponding ontology structure; Knowledge ontology module: includes IM, PM, and FM multi-dimensional information ontologies, which are used to uniformly store monitoring data, design parameters, manufacturing status, and simulation results of models from the physical space.

[0007] Preferably, the mapping mechanism in step S3 includes inheritance mapping: ; IM 1O Indicates 1 Information Model IM 1 The actual input information contained in IM 2 I Indicates 2 Information Model IM 2 The model virtual simulation information contained in the knowledge graph is represented as the attribute inheritance of the same entity, namely IM 1 The information output in O is consistent with IM 2 The information entered in I directly corresponds; this mapping is expressed as a direct relationship between "requirements analysis" → "conceptual design" without the need for additional knowledge reasoning.

[0008] Preferably, the mapping mechanism in step S3 includes aggregate mapping: ; IM i O Indicates i Information Model IM i The model virtual simulation information contained in the knowledge graph integrates the multi-entity information, that is, the data from multiple information sources are combined through aggregation relationships to form a new information model input.

[0009] Preferably, the mapping mechanism in step S3 includes adding a mapping: ; Adding a mapping is represented in the knowledge graph as an attribute extension of the entity, that is, the output information cannot fully meet the needs of the target twin model, and additional information A needs to be supplemented; adding a mapping is represented by associating "optimized design results" with "user needs", and a complete input information model is generated through knowledge supplementation.

[0010] Preferably, the mapping mechanism in step S3 includes extracting mapping: ; IM 1 Only some data items of the information model output in I are IM 2 Required by the information model input in I; Representation Extraction IM 1 O Specific data items in a i common k In the knowledge graph, extraction mapping is performed through "attribute screening", that is, extracting specific data items as information input for the digital twin model.

[0011] Preferably, step S4 integrates the reverse design optimization process of the knowledge graph and the digital twin, and the specific steps are as follows: S41: Define the initial one-way design process as , It is an approximate parameter obtained by analyzing the design requirements or empirical data. As variables related to system conditions and status in the input information model IMI, The data is transmitted to PM through the mapping process of the knowledge graph; in PM, the simulation model is called to perform virtual analysis on the structure to obtain performance prediction information. ,Will Input into FM to realize design parameters One-way optimization; S42: In the assumption When it is known a priori, the optimization algorithm can be used to solve make The optimal product design solution X* is obtained, because It cannot fully and accurately reflect the actual situation of the product and the design requirements, resulting in Deviations from the final design performance of the design solution; S43: Define the reverse design model as: ; Represents the actual response Y obtained through knowledge graph query reasoning IMI With model prediction The deviation between Indicates the deviation of the parameter The modified optimization function; the specific reverse process is: the actual response Y of the information model IMI IMI Feedback to the knowledge graph, through ontology reasoning and rule engine to mine relevant potential relationships and abnormal information, and continuously update , based on the actual system parameters ;Will As new design knowledge to replace the parameters in the original one-way mapping process ,Right now , thus forming a closed-loop process of reverse feedback of design decisions, A reverse design system for electromechanical products integrating knowledge graph and digital twin, comprising a data acquisition unit, a data processing unit and a data output unit; The data acquisition unit acquires the physical data and real-time operation data of the electromechanical products; The data processing unit combines the structured knowledge management capabilities of the knowledge graph KG with the dynamic update characteristics of the digital twin DT, so that the operation knowledge accumulated during the actual use of electromechanical products can be fed back to the design end in real time, forming a reverse feedback path for the continuous evolution of knowledge; Data output unit: Users can choose visual output at different stages according to their needs.

[0012] Preferably, the digital twin DT in the data processing unit builds a virtual model that is updated synchronously with the physical product, making the flow of data and the evolution of the model possible during the design process; the knowledge graph DT can use the real-time operation data collected by sensors to dynamically correct the virtual model, thereby accurately reflecting the state of the physical system.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention provides a method and system for reverse design of electromechanical products that integrates knowledge graphs and digital twins. By constructing a semantically driven digital twin model, the semantic unification and dynamic mapping of multi-source heterogeneous data throughout the entire life cycle are achieved, thereby enhancing the reliability and adaptability of the model in the product design, manufacturing, and operation and maintenance processes.

[0014] 2. The effective implementation of the present invention can realize the deep integration of multi-dimensional data information and knowledge reasoning, so that the operating knowledge accumulated during the actual use of electromechanical products can be fed back to the design end in real time, forming a reverse feedback path for the continuous evolution of knowledge, and improving the accuracy and foresight of product design.

[0015] 3. The unified semantic modeling framework and knowledge ontology structure proposed in the method of the present invention have good reusability and scalability, can avoid repeated design domain modeling and knowledge parsing tasks, ensure consistent expression and efficient interaction of data between multiple stages and multiple models, and improve the versatility and reusability of model modeling. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a principle framework diagram of the KG-based DT model information management mechanism invented in the present disclosure; Figure 2 This is a principle framework diagram of the KG-based information-model mapping mechanism invented in the present disclosure; Figure 3 A reverse design optimization process integrating KG and DT according to the disclosed invention; Figure 4 A reverse design optimization flow chart of the fusion of KG and DT for case introduction invented in the present disclosure; Figure 5 This is a main body structural diagram of the KG portion of the high-speed train bogie invented in the present invention for case introduction. DETAILED DESCRIPTION

[0017] The following detailed descriptions are exemplary and are intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present disclosure belongs.

[0018] In order to solve the problems of inconsistent formats and semantics of multi-source data, difficulty in dynamically adjusting model parameters, and difficulty in mapping the association between models and data in the traditional design process, a reverse design method and system for electromechanical products integrating knowledge graph and digital twin were proposed. The method and system consisted of 5 parts: S1 construction of digital twin model with multi-dimensional information association, S2 DT model information management mechanism based on KG, S3 information-model mapping mechanism based on KG, and S4 reverse design optimization process integrating KG and DT.

[0019] S1. Construction of digital twin model with multi-dimensional information association: The digital twin model is divided into three dimensions: information model, mechanism model and domain model. The metamodels corresponding to the three together constitute a complete digital twin model DTM. DTM is a digital representation of a physical entity. This includes its characteristics, attributes, transmission status, parameters, data and transmission capabilities. The mathematical expression of the digital twin model defined in the present invention is shown in the following formula.

[0020] .

[0021] It specifically includes the following three parts: S11: IM refers to a logical data set formed by organizing and structuring raw data according to a certain information modeling framework and semantic rules, aiming to achieve virtual representation and process tracking of physical objects. According to its different functional positions in the DTM model, IM can be divided into the following three types.

[0022] The input information model IMI in the digital twin model, the output information model IMO, and the twin information model IMDT generated during the digital twin model mapping process.

[0023] IMI is the input of DT model simulation and mapping, including product structure parameters, boundary conditions, load distribution, environmental variables, material properties and other information related to the physical system. IMO represents the virtual simulation prediction results and system response information obtained in the digital space, such as displacement, stress, modal frequency, temperature rise data, operating status feedback, etc. IMDT is the information data generated by DT during the mapping and feedback process. Such as intermediate variables in the model identification process, mapping associations, adjustment records in the optimization process, model correction results, etc.

[0024] S12: PM represents the mechanism model in the twin model. PM is a quantitative description of the physical behavior and internal mechanism of the product or system. PM models actual physical phenomena (such as structural stress, heat conduction, fluid flow, electromagnetic effects, etc.) and abstracts physical laws into mathematical expressions, such as partial differential equations, state space models, continuous mechanics models, etc. It is further transformed into a computational model that can be used for simulation analysis, such as the finite element method (FEM), multi-body dynamics (MBD), computational fluid dynamics (CFD) and other numerical computational models.

[0025] S13: FM represents the domain model in the digital twin model. FM is a collection of models built for the different behavioral characteristics and engineering application requirements of equipment throughout its life cycle. FM covers multiple sub-fields such as design optimization, operation monitoring, status assessment, life prediction, fault diagnosis, safety warning, etc. Each domain model corresponds to a specific life cycle stage and engineering task. For example, in the design stage, FM can integrate models such as structural optimization models and parameter sensitivity analysis to explore the design space and quickly find the optimal solution to achieve product optimization design.

[0026] S2. DT model information management mechanism based on KG: In the DT model, the interaction between the information model (IM), mechanism model (PM) and domain model (FM) determines the mapping method of information flow. The knowledge graph (KG) helps to achieve information exchange in the process of product design, manufacturing and operation by accessing, integrating and converting multi-dimensional information.

[0027] Since multidimensional information covers multiple key stages of the product life cycle, including geometric design, material selection, processing technology, and assembly process, etc. Therefore, it is necessary to first clarify the mapping direction of the information flow, determine the type of exchanged information, and classify the generated and applied information, such as geometric parameters, material properties, processing technology data, etc. The carriers of design, processing and operation information include text, images, and files in specific formats. Taking into account the diverse sources of knowledge, heterogeneous data formats, large amounts of data, and uneven quality, the DT model is combined with KG to manage multidimensional information and map and transmit between IM, PM, and FM sub-models, such as Figure 1 shown.

[0028] The KG-based DT model information management mechanism mainly includes the following three modules.

[0029] Data processing module: responsible for processing physical space data in different formats. This module uses a variety of data conversion and preprocessing technologies, including rule-based structured analysis, unstructured text information extraction based on natural language processing (NLP), sensor data normalization and time series interpolation processing, and multi-source data fusion algorithms (such as principal component analysis PCA and multimodal alignment technology) to normalize multi-source heterogeneous data into a standard format that can be parsed by knowledge graphs, thereby improving the accuracy and efficiency of subsequent knowledge extraction, entity recognition, and relationship modeling.

[0030] Key technology modules: including key technologies such as entity recognition and ambiguity resolution, which automatically fill instance information into the corresponding ontology structure. Named entity recognition models such as BiLSTM-CRF are used to extract entities from text and structured data, and contextual semantic embedding (such as BERT) is combined to improve the context adaptability of recognition. At the same time, an ambiguity resolution mechanism based on vector space model and semantic similarity calculation is introduced to solve the problems of homonymy and cross-domain entity confusion. The extraction results are structurally mapped and semantically verified through ontology matching and rule reasoning technology to ensure the consistency and integrity of data and ontology, so as to realize the structured expression and dynamic update of knowledge in the design, manufacturing and operation stages of the physical world.

[0031] Knowledge ontology module: including multi-dimensional information ontologies such as IM (information model), PM (mechanism model), FM (domain model), etc., which are used to uniformly store monitoring data, design parameters, manufacturing status and simulation results of models from physical space. For example, the stress response, temperature rise data and design performance evaluation results of structural parts can all be stored in the corresponding ontology instance. This module uses OWL language to construct the ontology architecture, uses Protégé for modeling and visualization management, and combines ontology alignment technology (such as structure matching algorithm based on semantic similarity) to achieve structural docking and unified expression between multi-source knowledge. At the same time, SPARQL is used for efficient data query, and SWRL rules are integrated to realize ontology reasoning and dynamic knowledge expansion, thereby improving the interpretability and flexibility of the knowledge graph.

[0032] S3. KG-based information-model mapping mechanism: The product life cycle involves the interaction and sharing of multi-dimensional information, which affects the construction of the twin model. Therefore, it is necessary to analyze the information mapping process between models at different stages. Combining KG with the definition of the information model FM in S1, the mapping process can be expressed as an information-model mapping mechanism based on KG, such as Figure 2 As shown. It mainly includes the following four independent mapping methods.

[0033] Inheritance Mapping: ; In KG, it is represented as the inheritance of attributes of the same entity, namely IM1 The information output in O is consistent with IM 2 The information input in I directly corresponds. For example, the requirement information output by the requirement analysis DT model can be directly used as the input of the conceptual design DT model. In KG, this mapping is expressed as a direct relationship between "requirement analysis" → "conceptual design" without additional reasoning.

[0034] Aggregate Mapping: ; Multi-entity information integration in KG means that data from multiple information sources are combined through aggregation relationships to form new information model input. For example, the manufacturing model input information may be composed of knowledge from multiple design models (such as structural design, material design, and process planning). In this case, multiple information sources need to be integrated through aggregation relationships.

[0035] Add the mapping: ; Adding a mapping is manifested in KG as an attribute extension of the entity, that is, the output information cannot fully meet the needs of the target twin model, and additional information A needs to be supplemented. For example, the output information of the optimized design DT model may also need to be combined with the user's personalized needs to form a complete decision-making plan. In KG, this mapping is manifested as "optimized design results" associated with "user needs", and a complete input information model is generated through knowledge supplementation.

[0036] Extract the mapping: ; IM 1 Only some data items of the information model output in I are IM 2 The information model input in I is required. This is manifested in KG as information screening and simplification. For example, in the mapping process from the structural design DT model to the process design DT model, only part of the information of the part structure diagram needs to be input into the process design stage. In KG, this mapping can be performed through "attribute screening". That is, extract specific data items and make them the information input of the target DT model.

[0037] S4. Reverse design optimization process integrating KG and DT: Traditionally, this process usually assumes model parameters or predefines module configurations, which results in a lack of feedback in the constructed design domain model. This leads to the adaptability of the design results and the failure to achieve optimal configuration of user needs. Therefore, based on the construction and mapping of the KG multidimensional twin model, this paper proposes Figure 3 The reverse design process of integrating KG and DT is shown. KG is used to manage and reason knowledge, and hidden design knowledge and demand information are mined through data feedback to achieve continuous iteration and upgrading of products. The reverse design process is as follows.

[0038] S41: Define the initial one-way design process as , the system parameters As variables related to system conditions and states in the input information model IMI, such as load distribution, boundary conditions, material properties, etc. required in structural design. The KG is mapped and transferred to the mechanism model PM. In PM, simulation models (such as ANSYS, Abaqus, etc.) are called to perform virtual analysis on the structure to obtain performance prediction information such as stress, displacement, and mode. .Will Input into the design domain model FM to realize the design parameters One-way optimization.

[0039] S42: In S41, Usually, approximate parameters are obtained by analyzing design requirements or empirical data, such as through DSE space exploration, combined with DoE experimental design methods, to construct the initial range of parameters under limited samples. Approximate parameters are estimated using regression analysis based on historical cases or machine learning prediction models (such as SVR, random forest and other technical methods). Then assume When the problem is known a priori, intelligent optimization techniques such as particle swarm optimization (PSO) and genetic algorithm (GA) are used to solve make The optimal product design is X*. It cannot fully and accurately reflect the actual situation of the product and the design requirements. There is a deviation from the design performance finally shown by the design solution.

[0040] S43: Therefore, based on the mapping of DT models in S1~S3 and the information management of KG, reverse design is defined as In the formula, Represents the actual response Y obtained through knowledge graph query reasoning IMI With model prediction The deviation between Indicates the deviation of the parameter The corrected optimization function. The specific reverse process is: the actual response Y of the information model IMI IMI Feedback to KG, through KG ontology reasoning (such as classification reasoning based on description logic, case reasoning) and rule engine (such as SWRL rules), relevant potential relationships and abnormal information are mined, combined with Bayesian inversion, Kalman filtering, particle filtering and other correction methods to continuously update , based on the actual system parameters .Will As new design knowledge to replace the parameters in the original one-way mapping process ,Right now , thus forming a closed-loop process of reverse feedback of design decisions.

[0041] S5. Case study on high-speed train bogies: The bogie of a high-speed train is a large and complex structure with large structural dimensions, extremely complex shapes, and high safety margins. The components of this equipment are large in size, and the interaction between them is complex, making the design extremely difficult. One of the focuses of bogie design is to optimize its frame structure, reduce the weight while ensuring strength, and achieve low-carbon and low-cost transportation. Among them, lightweight performance and strength performance are modeled as objective functions.

[0042] Due to the complexity of the bogie, parameters such as load and material strength are usually random and uncertain. This requires making assumptions based on prior experience or setting a high safety factor for the structural design, which leads to design redundancy. At the same time, if the feedback analysis of actual operating data is ignored, the natural frequency of the structure may be within the actual operating frequency range, and there is a risk of serious safety accidents during resonance. To this end, it is necessary to mine actual data information and feed it back to the design end to form a reverse design to improve the adaptability and reliability of the design. The case implementation steps are as follows: Figure 4 shown.

[0043] S51: First, it is necessary to manage the various parameters of the bogie, including geometric dimensions, material properties, boundary conditions, vibration response data and other multi-dimensional information, and build a multi-dimensional information knowledge ontology based on KG according to the composition of the DT model in S1. It is necessary to identify the multi-dimensional information data contained in the information model IMI, the design domain model FM, and the mechanism model PM, which covers the design scheme, response data, geometric shape, boundary conditions, modeling parameters, etc. of the bogie.

[0044] In order to effectively manage this information, named entity recognition (NER) technology based on rules and statistical learning is used to extract key entities from unstructured text and sensor data. Semantic alignment is performed in conjunction with the ontology dictionary, and a relationship extraction algorithm (such as template matching or BERT semantic relationship classifier) ​​is used to establish semantic connections between entities. Based on description logic rules and OWL semantic models, the three types of knowledge instances extracted are automatically filled into the corresponding FM, IMI and PM ontologies. The constructed ontology structure is shown in the figure below. Figure 5 shown.

[0045] S52: FM ontology includes the dimensional parameters of structures such as the crossbeam, side beam, inner plate of the bogie. IMI knowledge ontology includes the processing technology, acceleration monitoring signal, and force monitoring signal of the bogie. PM knowledge ontology includes parameters such as density, elastic stiffness, and damping of the model. After data processing and knowledge extraction steps, relevant instances are extracted and stored in the corresponding ontology. All structured knowledge is imported into the Neo4j graph database to build a multidimensional knowledge graph for the bogie.

[0046] S53: Strength performance Y 1 and lightweight performance 2 is the objective function. Based on the finite element theory, a mechanical analysis model of the high-speed train bogie is established to provide a mechanism model for structural optimization. Although numerical simulation can reflect its mechanical properties to a certain extent, it is necessary to assume system parameters based on empirical estimates or predefined ranges. , including density, elastic stiffness and damping parameters. These parameters are important factors affecting the performance of high-speed train bogies. The system parameters are corrected according to actual data. This is the unique advantage of reverse design. Replace the assumed , which helps product improvement and design optimization.

[0047] Through the KG in S52, multi-dimensional information management is performed and mapped to the mechanism model FM. By developing an intelligent correction algorithm, parameters such as density, elastic stiffness and damping are updated to build a twin IMDT model consistent with the actual output. At this time, more accurate prediction values ​​of target performance such as strain and mass can be obtained. , to guide the optimization of the bogie mechanism, that is, . Map the optimization solution to the design domain model, and further iteratively optimize the structure according to actual design requirements. Figure 4 The actual train operation data is analyzed to obtain the operating frequency range of the bogie. The strength and volume design schemes are coupled, and the natural frequency of the bogie is taken as the optimization target. 3 Perform secondary optimization. This process maps the physical data of the bogie to the DT model based on KG, and continuously feeds back the operation data to the design domain model to form a reverse design to enhance the continuous optimization iteration of the bogie.

[0048] A reverse design system for electromechanical products integrating knowledge graph and digital twin, comprising a data acquisition unit, a data processing unit and a data output unit; The data acquisition unit acquires the physical data and real-time operation data of the electromechanical products; The data processing unit combines the structured knowledge management capabilities of the knowledge graph KG with the dynamic update characteristics of the digital twin DT, so that the operation knowledge accumulated during the actual use of electromechanical products can be fed back to the design end in real time, forming a reverse feedback path for the continuous evolution of knowledge; The digital twin DT in the data processing unit builds a virtual model that is updated synchronously with the physical product, making the flow of data and the evolution of the model possible during the design process; the knowledge graph DT can use the real-time operation data collected by sensors to dynamically modify the virtual model, thereby accurately reflecting the status of the physical system; Data output unit: Users can choose visual output at different stages according to their needs.

[0049] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A reverse design method for electromechanical products integrating knowledge graph and digital twin, characterized in that: The following steps are involved: S1. Construct a digital twin model with multi-dimensional information association; The digital twin model is divided into three dimensions: information model, mechanism model and domain model. The metamodels corresponding to the three together constitute a complete digital twin model DTM; S2. Establish a digital twin model information management mechanism based on knowledge graph; In the digital twin model, the interaction between the information model IM, the mechanism model PM and the domain model FM determines the mapping method of the information flow; the knowledge graph helps to achieve information exchange in the product design, manufacturing and operation process by accessing, integrating and converting multi-dimensional information; S3. Establish an information-model mapping mechanism based on knowledge graph; Analyze multi-dimensional information data at different stages of the product life cycle, use knowledge graphs to uniformly manage multi-source heterogeneous information, and establish multiple semantic mapping relationships between information and models; S4. Reverse design optimization process integrating knowledge graph and digital twin; The knowledge graph is used for querying and reasoning about knowledge. It updates the digital twin model by mining hidden design knowledge and demand information through data association, forming a continuous feedback knowledge update loop and enhanced reverse optimization design.

2. The electromechanical product reverse design method integrating knowledge graph and digital twin according to claim 1 is characterized in that: In step S1, the digital twin model expression is as follows: ； Information model IM refers to a logical data set formed by organizing and structuring the original data according to a certain information modeling framework and semantic rules; among them: IMI is the input of digital twin model simulation and mapping, IMO represents the virtual simulation prediction results and system response information obtained in the digital space, and IMDT is the information data generated by the digital twin during the mapping and feedback process; The mechanism model PM is a quantitative description of the physical behavior and internal mechanism of a product or system; The domain model FM is a collection of models built based on the different behavioral characteristics and engineering application requirements of equipment throughout its life cycle.

3. The electromechanical product reverse design method integrating knowledge graph and digital twin according to claim 1 is characterized in that: The digital model information management mechanism based on knowledge graph established in step S2 includes the following three modules: Data processing module: responsible for processing physical space data in different formats; The key technology module fills the instance information into the corresponding ontology structure; Knowledge ontology module: includes IM, PM, and FM multi-dimensional information ontologies, which are used to uniformly store monitoring data, design parameters, manufacturing status, and simulation results of models from the physical space.

4. The electromechanical product reverse design method integrating knowledge graph and digital twin according to claim 1 is characterized in that: The mapping mechanism in step S3 includes inheritance mapping: ； IM 1 O Indicates 1 Information Model IM 1 The actual input information contained in IM 2 I Indicates 2 Information Model IM 2 The model virtual simulation information contained in IM1O is represented as the attribute inheritance of the same entity in the knowledge graph, that is, the information output in IM1O directly corresponds to the information input in IM2I.

5. The electromechanical product reverse design method integrating knowledge graph and digital twin according to claim 1 is characterized in that: The mapping mechanism in step S3 includes aggregate mapping: ； IM i O Indicates i Information Model IM i The model virtual simulation information contained in the knowledge graph integrates the multi-entity information, that is, the data from multiple information sources are combined through aggregation relationships to form a new information model input.

6. The electromechanical product reverse design method integrating knowledge graph and digital twin according to claim 1 is characterized in that: The mapping mechanism in step S3 includes adding a mapping: ； Adding a mapping is represented in the knowledge graph as an attribute extension of the entity, that is, the output information cannot fully meet the needs of the target twin model, and additional information A needs to be supplemented; adding a mapping is represented by associating "optimized design results" with "user needs", and a complete input information model is generated through knowledge supplementation.

7. The electromechanical product reverse design method integrating knowledge graph and digital twin according to claim 1 is characterized in that: The mapping mechanism in step S3 includes extracting the mapping: ； Representation Extraction IM 1 O Specific data items in a i common k items and use them as information input for the digital twin model.

8. The electromechanical product reverse design method integrating knowledge graph and digital twin according to claim 1 is characterized in that: Step S4 integrates the reverse design optimization process of knowledge graph and digital twin. The specific steps are as follows: S41: Define the initial one-way design process as , It is an approximate parameter obtained by analyzing the design requirements or empirical data. As variables related to system conditions and status in the input information model IMI, Transmitted to PM through the knowledge graph mapping process; In PM, the simulation model is called to perform virtual analysis on the structure to obtain performance prediction information ,Will Input into FM to realize design parameters One-way optimization; S42: In the assumption When it is known a priori, the optimization algorithm can be used to solve make The optimal product design solution X* is obtained, because It cannot fully and accurately reflect the actual situation of the product and the design requirements, resulting in Deviations from the final design performance of the design solution; S43: Define the reverse design model as: ； Represents the actual response Y obtained through knowledge graph query reasoning IMI With model prediction The deviation between Indicates the deviation of the parameter Corrected optimization function; The specific reverse process is: the actual response Y of the information model IMI IMI Feedback to the knowledge graph, through ontology reasoning and rule engine to mine relevant potential relationships and abnormal information, and continuously update , based on the actual system parameters ;Will As new design knowledge to replace the parameters in the original one-way mapping process ,Right now , thus forming a closed-loop process of reverse feedback of design decisions.

9. A reverse design system for electromechanical products integrating knowledge graph and digital twin, using the reverse design method for electromechanical products integrating knowledge graph and digital twin as described in any one of claims 1 to 8, characterized in that: It includes a data acquisition unit, a data processing unit and a data output unit; The data acquisition unit acquires the physical data and real-time operation data of the electromechanical products; The data processing unit combines the structured knowledge management capabilities of the knowledge graph KG with the dynamic update characteristics of the digital twin DT, so that the operation knowledge accumulated during the actual use of electromechanical products can be fed back to the design end in real time, forming a reverse feedback path for the continuous evolution of knowledge; Data output unit: Users can choose visual output at different stages according to their needs.

10. The electromechanical product reverse design system integrating knowledge graph and digital twin according to claim 9 is characterized in that: The digital twin DT in the data processing unit builds a virtual model that is updated synchronously with the physical product, making the flow of data and the evolution of the model possible during the design process; the knowledge graph DT can use the real-time operation data collected by sensors to dynamically correct the virtual model, thereby accurately reflecting the state of the physical system.

Images