Knowledge graph-based production and manufacturing digital twin modeling method and device, and electronic equipment

Abstract

The application relates to the technical field of knowledge engineering, and discloses a production and manufacturing digital twin modeling method based on a knowledge graph, which comprises the following steps: performing data fusion and model construction on original multi-source data, obtaining multi-source fusion data, and generating a digital twin architecture; performing ontological design and semantic modeling on the multi-source fusion data, and generating a manufacturing knowledge graph; wherein the manufacturing knowledge graph is represented by triplets, and the triplets comprise entities-relation-entities / attributes; performing entity modeling on the manufacturing knowledge graph, and obtaining a knowledge-driven geometric model; and performing reasoning based on the manufacturing knowledge graph and historical data, and obtaining a knowledge-driven decision model. The method can construct a digital twin model with knowledge-driven prediction and decision-making capabilities, and provides a high-robustness and sustainable-evolution digital twin modeling method for intelligent manufacturing, smart factories and other fields. The application also discloses a production and manufacturing digital twin modeling device based on the knowledge graph and an electronic device.

Description

Technical Field

[0001] This application relates to the field of knowledge engineering technology, such as a knowledge graph-based digital twin modeling method and apparatus for manufacturing, and electronic equipment. Background Technology

[0002] Currently, digital twin technology was first proposed by NASA. Its core idea is to map the operational state of physical objects into virtual models, achieving a combination of virtual and real elements and dynamic interaction. In recent years, with the development of the Internet of Things, artificial intelligence, and big data, digital twins have gradually become an important foundation for fields such as intelligent manufacturing, smart cities, and energy and power. Especially in manufacturing, digital twins, through geometric modeling and physical simulation, can provide reliable support for product design, process planning, and equipment maintenance. However, existing digital twin research has shortcomings in the following aspects:

[0003] 1. Data isolation and semantic inconsistency: The manufacturing process involves multi-source heterogeneous data, including real-time status data collected by sensors, product CAD (Computer-Aided Design) models, process parameters, historical cases, and equipment logs. These data often exist in different formats and modalities, lacking a unified semantic modeling approach, making it difficult to achieve effective data fusion and retrieval when building twin models.

[0004] 2. Lack of a knowledge-driven mechanism: Currently, most digital twin models still rely on traditional physical modeling and numerical calculation methods. While they can reflect certain physical behaviors, their utilization of knowledge such as process rules, expert experience, and historical optimization cases is limited. Existing models often remain at the level of "data-driven + physical simulation," making it difficult to form a mechanism for autonomous learning, dynamic reasoning, and knowledge updating.

[0005] 3. Insufficient model evolution capability: Digital twin objects continuously accumulate new observation data and experience during operation, but existing methods lack effective knowledge updating and evolution mechanisms. Over time, the twin is prone to "knowledge aging," causing the virtual model to gradually deviate from the real system.

[0006] 4. Lack of Closed-Loop Optimization: In actual manufacturing processes, there are complex factors such as discrepancies between planning and execution, and fluctuations in process conditions. Existing twin models mostly remain at the virtual simulation stage, making it difficult to provide timely optimization suggestions or control commands to the physical entity. In other words, the twin lacks a closed-loop mechanism of "perception-reasoning-simulation-feedback," and cannot effectively guide the dynamic optimization of the manufacturing system.

[0007] 5. Limited versatility and scalability: Existing digital twins are mostly dedicated modeling methods for specific processes or equipment, lacking cross-domain abstraction and knowledge transfer capabilities. When the application scenario or equipment type changes, remodeling is often required, which is costly and inefficient.

[0008] To overcome the above problems, scholars and enterprises at home and abroad have proposed some improvement strategies, such as: realizing multi-source data collection and fusion through IoT platforms; introducing big data analysis methods to uncover potential patterns; and using machine learning models for simulation prediction and parameter optimization.

[0009] However, due to the lack of a unified knowledge modeling framework and cross-domain semantic standards, these improvements often remain at the level of local optimization and have not yet formed a systematic approach. Especially in the manufacturing industry, how to use knowledge graph and ontology modeling techniques to formally represent multi-source manufacturing knowledge and drive digital twins to achieve self-evolution and optimization decisions remains a pressing technical challenge. Summary of the Invention

[0010] To provide a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. This summary is not intended as a general commentary, nor is it intended to identify key / important components or describe the scope of protection of these embodiments, but rather as a prelude to the detailed description that follows.

[0011] This disclosure provides a knowledge graph-based digital twin modeling method, apparatus, and electronic device for manufacturing, to construct a digital twin model with knowledge-driven prediction and decision-making capabilities, providing a highly robust and sustainably evolving digital twin modeling method for multiple fields such as intelligent manufacturing and smart factories.

[0012] In some embodiments, the method includes: performing data fusion and model building on raw multi-source data to obtain multi-source fused data and generate a digital twin architecture; performing ontology design and semantic modeling on the multi-source fused data to generate a manufacturing knowledge graph; wherein the manufacturing knowledge graph is represented by triples, and the triples include entity-relationship-entity / attribute; performing entity modeling on the manufacturing knowledge graph to obtain a knowledge-driven geometric model; and performing reasoning based on the manufacturing knowledge graph and historical data to obtain a knowledge-driven decision model.

[0013] In some embodiments, ontological design and semantic modeling are performed on multi-source fusion data to generate a manufacturing knowledge graph, including: sequentially performing entity recognition and relation extraction, and attribute mapping on the multi-source fusion data to obtain triples represented by entity-relationship-entity / attribute; and constructing a manufacturing knowledge graph based on the triples.

[0014] In some embodiments, entity modeling is performed on the manufacturing knowledge graph to obtain a knowledge-driven geometric model, including: obtaining system attribute information of the manufacturing system and constructing semantic rules of the manufacturing knowledge graph based on the manufacturing knowledge graph; wherein, the attribute information includes component information and assembly relationships, process constraints; establishing a mapping relationship between key dimensions of geometric entities and knowledge parameters based on component information; generating virtual object models corresponding to the geometric entities based on the semantic rules and mapping relationships of the manufacturing knowledge graph; and constructing a geometric model of the virtual space based on the virtual object models.

[0015] In some embodiments, entity modeling of the manufacturing knowledge graph to obtain a knowledge-driven geometric model further includes: feature extraction and registration of point cloud data to align with the constructed geometric model, and model correction of the virtual object model.

[0016] In some embodiments, reasoning based on a manufacturing knowledge graph and historical data is used to obtain a knowledge-driven decision-making model, including: retrieving and reasoning based on the manufacturing knowledge graph to obtain production logic; wherein the production logic includes resource constraints and process rules, and scheduling logic; performing semantic parsing on the production logic to construct a reasoning logic library; wherein the reasoning logic library is used to call the decision-making model; mapping key production data to semantic nodes of the manufacturing knowledge graph, and fusing the production data with the manufacturing knowledge graph; wherein the key production data includes equipment operating status and production progress, process parameters, and environmental information; and constructing a decision-making model in a virtual space based on the semantic rules of the reasoning logic library and the manufacturing knowledge graph.

[0017] In some embodiments, the original multi-source data includes geometric measurement data, equipment status data, environmental monitoring data, and process record data. Data fusion and model building are performed on the original multi-source data to obtain multi-source fused data and generate a digital twin architecture. This includes: classifying the original multi-source data to obtain multi-source classified data; sequentially cleaning and removing anomalies, aligning the time sequence, and unifying the format of the multi-source classified data to obtain multi-source fused data; and constructing the digital twin architecture. The digital twin architecture includes a physical layer, a data layer, and a knowledge-driven virtual layer. The physical layer is used to collect the operating status of equipment and processes, the data layer is used to aggregate, manage, and process the original multi-source data, and the knowledge-driven virtual layer has semantic association and reasoning capabilities based on a manufacturing knowledge graph.

[0018] In some embodiments, the method further includes: after obtaining a knowledge-driven decision model by reasoning based on the manufacturing knowledge graph and historical data, embedding a knowledge-enhanced prediction model (KGE) into the decision model to perform semantic enhancement and structural constraints on the manufacturing knowledge graph before model prediction.

[0019] In some embodiments, the apparatus includes: a virtual architecture construction module configured to perform data fusion and model building on raw multi-source data to obtain multi-source fused data and generate a digital twin architecture; a knowledge graph construction module configured to perform ontological design and semantic modeling on the multi-source fused data to generate a manufacturing knowledge graph; wherein the manufacturing knowledge graph is represented by triples, and the triples include entity-relationship-entity / attribute; a geometry construction module configured to perform entity modeling on the manufacturing knowledge graph to obtain a knowledge-driven geometric model; and a knowledge reasoning module configured to perform reasoning based on the manufacturing knowledge graph and historical data to obtain a knowledge-driven decision model.

[0020] In some embodiments, the apparatus includes a processor and a memory storing program instructions, the processor being configured to execute, when running the program instructions, the knowledge graph-based digital twin modeling method for manufacturing described above.

[0021] In some embodiments, the electronic device includes: a device body; and a knowledge graph-based digital twin modeling device for manufacturing, as described above, installed on the device body.

[0022] The knowledge graph-based digital twin modeling method, apparatus, and electronic device for manufacturing provided in this disclosure can achieve the following technical effects:

[0023] This disclosure embodiment performs data fusion and model building on original multi-source data to obtain multi-source fused data and generate a digital twin architecture. Then, ontological design and semantic modeling are performed on the multi-source fused data to generate a manufacturing knowledge graph. Entity modeling is then performed on the manufacturing knowledge graph to obtain a set model, and reasoning is performed based on the manufacturing knowledge graph and historical data to obtain a knowledge-driven decision model. In this disclosure embodiment, the manufacturing knowledge graph is used to semantically express and reason about equipment, parts, processes, and their relationships in the manufacturing system. This not only improves data relevance and interpretability but also enables the digital twin architecture to possess knowledge-driven prediction and decision-making capabilities. Simultaneously, by directly applying intelligent decisions from the virtual space to the physical system through the decision model, system operating efficiency is effectively improved, and system robustness and adaptability are enhanced. Therefore, a highly robust and sustainably evolving digital twin modeling method is provided for multiple fields such as intelligent manufacturing and smart factories.

[0024] The above general description and the description below are exemplary and illustrative only and are not intended to limit this application. Attached Figure Description

[0025] One or more embodiments are illustrated by way of example with reference to the accompanying drawings. These illustrations and drawings do not constitute a limitation on the embodiments. Elements having the same reference numerals in the drawings are shown as similar elements. The drawings are not to be scaled. And wherein:

[0026] Figure 1 This is a schematic diagram of a knowledge graph-based digital twin modeling method for manufacturing provided in an embodiment of this disclosure;

[0027] Figure 2 This is a schematic diagram of a model flow for a knowledge graph-based digital twin modeling method for manufacturing provided in this embodiment of the disclosure;

[0028] Figure 3 This is a schematic diagram of another knowledge graph-based digital twin modeling method for manufacturing provided in this disclosure embodiment;

[0029] Figure 4 This is a schematic diagram of another knowledge graph-based digital twin modeling method for manufacturing provided in this disclosure embodiment;

[0030] Figure 5 This is a schematic diagram of another knowledge graph-based digital twin modeling method for manufacturing provided in this disclosure embodiment;

[0031] Figure 6 This is a schematic diagram of another knowledge graph-based digital twin modeling method for manufacturing provided in this disclosure embodiment;

[0032] Figure 7 This is a schematic diagram illustrating the construction of a knowledge-driven geometric model provided in an embodiment of this disclosure;

[0033] Figure 8 This is a schematic diagram illustrating the construction of the knowledge-driven decision-making model provided in this embodiment of the disclosure;

[0034] Figure 9 This is a schematic diagram of a knowledge graph-based digital twin modeling device for manufacturing provided in an embodiment of this disclosure;

[0035] Figure 10 This is a schematic diagram of another knowledge graph-based digital twin modeling device for manufacturing provided in this embodiment. Detailed Implementation

[0036] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this disclosure described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion.

[0037] Unless otherwise stated, the term "multiple" means two or more.

[0038] In this embodiment of the disclosure, the character " / " indicates that the objects before and after it are in an "or" relationship. For example, A / B means: A or B.

[0039] The term "and / or" describes an association between objects, indicating that three relationships can exist. For example, A and / or B means: A or B, or A and B.

[0040] The term "correspondence" can refer to an association or binding relationship. The correspondence between A and B means that there is an association or binding relationship between A and B.

[0041] Combination Figure 1 and Figure 2 As shown, this disclosure provides a knowledge graph-based digital twin modeling method for manufacturing, including:

[0042] S01, the processor performs data fusion and model building on the original multi-source data to obtain multi-source fused data and generate a digital twin architecture.

[0043] In this step, the raw multi-source data includes geometric measurement data, equipment status data, environmental monitoring data, and process record data.

[0044] S02, the processor performs ontology design and semantic modeling on the multi-source fused data to generate a manufacturing knowledge graph. The manufacturing knowledge graph is represented by triples, which include entity-relationship-entity / attribute. Entity-relationship-entity / attribute includes entity-relationship-entity, or entity-relationship-attribute.

[0045] In this step, the knowledge graph can be abstractly represented by the following formula:

[0046] .in, For raw multi-source data, For triplet geometry, The number of original multi-source data entries. The attribute dimensions of the original multi-source data. The number of triples is 3, which corresponds to the basic structure of entity-relationship-entity / attribute.

[0047] S03, the processor performs entity modeling on the manufacturing knowledge graph to obtain a knowledge-driven geometric model.

[0048] S04, the processor performs reasoning based on the manufacturing knowledge graph and historical data to obtain a knowledge-driven decision model.

[0049] This disclosure employs a knowledge graph-based digital twin modeling method for manufacturing. This method fuses and constructs models from original multi-source data to obtain multi-source fused data and generate a digital twin architecture. The multi-source fused data is then subjected to ontological design and semantic modeling to generate a manufacturing knowledge graph. Entity modeling is performed on the manufacturing knowledge graph to obtain a set model, and reasoning is performed based on the manufacturing knowledge graph and historical data to obtain a knowledge-driven decision-making model. In this disclosure, the manufacturing knowledge graph is used to semantically represent and reason about equipment, parts, processes, and their relationships in the manufacturing system. This not only improves data relevance and interpretability but also enables the digital twin architecture to possess knowledge-driven prediction and decision-making capabilities. Simultaneously, by directly applying intelligent decisions from the virtual space to the physical system through the decision model, system operating efficiency is effectively improved, and system robustness and adaptability are enhanced. Therefore, this provides a highly robust and sustainably evolving digital twin modeling method for multiple fields such as intelligent manufacturing and smart factories.

[0050] Optionally, combined Figure 3 As shown, the processor performs ontology design and semantic modeling on multi-source fused data to generate a manufacturing knowledge graph, including:

[0051] The processor sequentially performs entity recognition, relation extraction, and attribute mapping on the multi-source fused data to obtain triples represented by entity-relationship-entity / attribute. Entities include equipment or components, processing steps, and monitoring links. Relationships include dependencies or constraints, combinations, and upstream / downstream connections. Attributes include geometric dimensions or accuracy indicators, operating parameters, and lifespan.

[0052] The processor constructs a manufacturing knowledge graph based on triples.

[0053] Thus, this embodiment of the disclosure constructs a clear conceptual hierarchy and attribute definition for different dimensions such as equipment, parts, processes, and environment, forming a unified semantic framework in conjunction with manufacturing domain standards. Specifically, it sequentially performs entity recognition and relation extraction, and attribute mapping on multi-source fused data to obtain triples represented by entity-relationship-entity / attribute. This embodiment of the disclosure also constructs an scalable manufacturing knowledge graph based on triples. By introducing entity recognition, relation extraction, and attribute mapping, it achieves knowledge merging and conflict resolution from different data sources and different abstraction levels, providing high-semantic quality support for the dynamic driving and optimization of subsequent digital twin architecture.

[0054] Optionally, combined Figure 3 As shown, the processor performs entity modeling on the manufacturing knowledge graph to obtain a knowledge-driven geometric model, including:

[0055] S11, the processor obtains the system attribute information of the manufacturing system based on the manufacturing knowledge graph and constructs the semantic rules of the manufacturing knowledge graph. The attribute information includes component information, assembly relationships, and process constraints. Component information includes system structure composition and structural hierarchy; assembly relationships include assembly sequence and key component associations; and process constraints include processing technology flow.

[0056] In this step, the processor obtains the system attribute information of the manufacturing system based on the manufacturing knowledge graph and constructs semantic rules for the manufacturing knowledge graph. This includes: the processor semantically expressing the component information, assembly relationships, and process constraints of the manufacturing system based on the manufacturing knowledge graph, and determining the spatial location, morphological features, and assembly dependencies of elements through logical reasoning. This provides semantic rules for geometric model construction.

[0057] S12, the processor establishes the mapping relationship between the key dimensions of the geometric entity and the knowledge parameters based on the component information, generates the virtual object model corresponding to the geometric entity based on the semantic rules and mapping relationship of the manufacturing knowledge graph, and constructs the geometric model of the virtual space based on the virtual object model.

[0058] Thus, in the digital twin architecture, the geometric model serves as the fundamental carrier for realizing virtual-real mapping, and knowledge-driven geometric model construction can establish a precise semantic and collaborative connection between the physical manufacturing system and the virtual model. To construct the geometric model for virtual space, the system attribute information of the manufacturing system is first obtained based on the manufacturing knowledge graph, and semantic rules for the manufacturing system are constructed. Then, based on component information, a mapping relationship between the key dimensions of geometric entities and knowledge parameters is established. Based on the semantic rules and mapping relationships of the manufacturing knowledge graph, virtual object models corresponding to the geometric entities are generated, and the geometric model of the virtual space is constructed based on these virtual object models. At this stage, the dimensional parameters, spatial positions, and kinematic constraints of each part in the manufacturing system are gradually injected through semantic rules, making the geometric model generation process no longer solely dependent on static CAD information, but possessing dynamic scalability and logical consistency. Furthermore, through the mapping relationship between the key dimensions of geometric entities and knowledge parameters, virtual object models at different precision levels can be generated in real time. Parametric modeling techniques can then be used to establish multi-resolution geometric expressions for virtual objects of different scales, from the overall to the local, ensuring that the geometric model always maintains strong semantic consistency with the actual manufacturing system. In this way, by leveraging knowledge-driven geometric model construction, not only is the efficiency and accuracy of geometric model generation improved, but the consistency of geometric model structure and logic can also be guaranteed, providing a traceable, evolvable and stable virtual foundation for subsequent production simulation, state prediction and optimization decision-making.

[0059] The above geometric model can be abstractly represented by the following formula: .in, A virtual object model corresponding to a geometric entity. For the key dimensions of the geometric solid, To create semantic rules for knowledge graphs A knowledge-driven geometry generation mechanism.

[0060] Optionally, the processor performs entity modeling on the manufacturing knowledge graph to obtain a knowledge-driven geometric model, and also includes: the processor extracts and registers features from point cloud data to align with the constructed geometric model, and performs model correction on the virtual object model.

[0061] Thus, this embodiment of the disclosure further performs feature extraction and registration on the point cloud data to align it with the constructed geometric model, thereby achieving model correction of the virtual object model. This mechanism ensures that the digital twin architecture not only maintains accuracy in the initial stage of model construction, but also achieves high-fidelity synchronization during the dynamic evolution of the production process.

[0062] The above model correction can be expressed by the following formula: .in, For a moment The constructed geometric model, This refers to the correction amount obtained through feature extraction and registration using point cloud data. This is the updated geometric model.

[0063] In practical applications, a closed-loop mechanism of knowledge-driven, parameter mapping, model building, and model correction is used to construct the geometric model, enabling it to respond in real time to changes in the manufacturing system. Combined with... Figure 7 As shown, the geometric model is no longer a static and isolated three-dimensional form, but a dynamic carrier that can integrate semantics, logical constraints and real-time perception, providing reliable model support for the construction of subsequent dining models and state prediction.

[0064] Optionally, combined Figure 4 As shown, the processor performs reasoning based on the manufacturing knowledge graph and historical data to obtain a knowledge-driven decision-making model, including:

[0065] S21, the processor retrieves and reasons based on the manufacturing knowledge graph to obtain the production logic. The production logic includes resource constraints and process rules, as well as scheduling logic.

[0066] S22, the processor performs semantic parsing on the production logic and constructs an inference logic library. This inference logic library is used to invoke the decision model.

[0067] S23, the processor maps key production data to semantic nodes in the manufacturing knowledge graph, fusing the production data with the manufacturing knowledge graph. Key production data includes equipment operating status and production progress, process parameters, and environmental information. Semantic nodes in the manufacturing knowledge graph represent entity nodes within triples.

[0068] S24, the processor constructs a decision model in virtual space based on the semantic rules of the reasoning logic library and the knowledge graph.

[0069] In this step, the decision model includes optimization schemes. These optimization schemes may include some or all of the resource reallocation strategies, production sequence adjustments, and parameter optimization strategies. After constructing the decision model in virtual space based on the semantic rules of the inference logic library and the manufacturing knowledge graph, the processor also includes outputting the optimization schemes through the system interface of the manufacturing system.

[0070] Thus, in the digital twin architecture, the construction of the decision model undertakes the task of transforming geometric and behavioral models into production optimization and management strategies. Its core idea is to leverage the semantic rules and reasoning logic provided by the manufacturing knowledge graph to fuse real-time and historical data from the manufacturing environment, thereby forming a multi-objective intelligent decision-making framework in virtual space. During the decision model construction process, the production logic is first obtained through retrieval and reasoning based on the manufacturing knowledge graph, and then semantically parsed to construct a reasoning logic library that can be called upon by the decision model. This mechanism ensures the inherent consistency and interpretability of the decision-making mechanism, enabling the decision model to not only rely on numerical calculation results but also ensure its rationality and traceability through knowledge-driven approaches.

[0071] During the decision-making model construction phase, key production data collected in real-time perception is mapped to semantic nodes of the manufacturing knowledge graph to fuse production data with the manufacturing knowledge graph. This enables the decision-making model to quickly identify conflicting factors and potential bottlenecks in the current production scenario. For example, when a processing delay is detected, the manufacturing knowledge graph can infer the cause of the delay based on built-in logic (such as insufficient resource allocation or unreasonable system parameter settings). The introduction of this logical reasoning gives the decision-making not only the advantages of quantitative analysis but also qualitative knowledge support. Finally, a decision-making model is constructed and generated in virtual space based on the reasoning logic library and the semantic rules of the manufacturing knowledge graph. Figure 8 This illustrates the construction of a knowledge-driven decision-making model. (Combined with...) Figure 8 As shown, the knowledge-driven decision-making model generated through the above mechanism has three characteristics: logic-driven, data-driven, and dynamic adaptation. This not only ensures the scientificity and accuracy of the modeling, but also significantly enhances the application value of digital twins in the manufacturing process, providing key support for subsequent state prediction and decision support.

[0072] The above decision-making model can be represented by the following formula: .in, The model output of the decision-making model. To retrieve and reason about the production logic obtained from the manufacturing knowledge graph. For a moment Key production data, This is a knowledge-driven reasoning and optimization function.

[0073] Optionally, combined Figure 5 As shown, the processor performs data fusion and model building on the original multi-source data to obtain multi-source fused data and generate a digital twin architecture, including:

[0074] S31, the processor performs data classification on the original multi-source data to obtain multi-source classified data.

[0075] In this step, the processor classifies the raw multi-source data to obtain multi-source classified data. This includes: the processor classifies the raw multi-source data according to structured data standards to obtain multi-source classified data. The multi-source classified data includes structured data, unstructured data, and semi-structured data. Structured data includes examples such as temperature, rotational speed, and machining accuracy; unstructured data includes examples such as expert experience documents, drawings, and log records; and semi-structured data includes examples such as process routes and operating specifications.

[0076] S32, the processor sequentially cleans and removes anomalies from the multi-source classification data, aligns the time sequence, and unifies the format to obtain multi-source fused data.

[0077] In this step, the processor sequentially cleans and removes anomalies from the multi-source classification data, aligns it temporally, and standardizes its format to obtain multi-source fused data. This includes: the processor performing rule parsing on the semi-structured data to extract processing steps and parameters, obtaining first structured intermediate data; and / or, the processor importing standardized field formats from the structured data to obtain second structured intermediate data. The processor then sequentially cleans and removes anomalies from the first structured intermediate data, the second structured intermediate data, and the structured data, aligns them temporally, and standardizes their formats to obtain multi-source fused data. The multi-source fused data is represented by a multidimensional vector dataset.

[0078] The S33 processor constructs a digital twin architecture. This digital twin architecture includes a physical layer, a data layer, and a knowledge-driven virtual layer. The physical layer is used to collect the operating status of equipment and processes; the data layer is used to aggregate, manage, and process raw multi-source data; and the knowledge-driven virtual layer possesses semantic association and reasoning capabilities based on a manufacturing knowledge graph.

[0079] Thus, in this embodiment, the heterogeneous original multi-source data is first classified to obtain multi-source classified data. Then, the multi-source classified data is sequentially cleaned, anomaly removed, time-series aligned, and formatted to obtain multi-source fused data, ensuring data consistency and validity. Based on this, a three-layer digital twin architecture, consisting of a physical layer, a knowledge-driven virtual layer, and a data layer, is constructed with data as the core driver to form a unified semantic framework conforming to manufacturing standards. The physical layer is responsible for collecting the operating status of equipment and processes; the knowledge-driven virtual layer undertakes semantic relationships and reasoning capabilities based on manufacturing knowledge graphs; and the data layer is responsible for the aggregation, management, and processing of the original multi-source data. These three layers map and interact with each other, forming a dynamic mapping relationship integrating the virtual and physical worlds, providing high-quality input and multi-level support for subsequent knowledge modeling, process optimization, and decision support.

[0080] Combination Figure 6 As shown in the embodiments of this disclosure, a method for digital twin modeling of manufacturing based on knowledge graphs is also provided, including:

[0081] S41, the processor performs data fusion and model building on the original multi-source data to obtain multi-source fused data and generate a digital twin architecture.

[0082] S42, the processor performs ontology design and semantic modeling on the multi-source fused data to generate a manufacturing knowledge graph. The manufacturing knowledge graph is represented by triples, which include entity-relationship-entity / attribute.

[0083] S43, the processor performs entity modeling on the manufacturing knowledge graph to obtain a knowledge-driven geometric model.

[0084] S44, the processor performs reasoning based on the manufacturing knowledge graph and historical data to obtain a knowledge-driven decision model.

[0085] S45, the processor embeds KGE (Knowledge Graph Embedding) into the decision model to semantically enhance and structurally constrain the manufactured knowledge graph before model prediction.

[0086] In this step, the knowledge enhancement prediction module can be a KGE (Knowledge Graph Embedding) module. This maps entities and relationships in the generated knowledge graph into low-dimensional dense vectors, which are then concatenated with or fused with attention from the original time-series features.

[0087] In some alternative embodiments, the knowledge graph embedding module is, for example, the TransE model. The specific formula is expressed as follows: .in, The embedding vector of the tail entity. The embedding vector of the head entity. The relation vectors are used. Additive approximations of relations in the vector space are used to maintain the consistency of the knowledge structure. This semantic embedding is used as additional feature input to introduce structural constraints from the manufacturing domain. Optionally, the TransE model employs the Adam (Adaptive Moment Estimation) optimizer. Optionally, the TransE model also models the semantic connections between entities and relations through translation operations, and its training process uses a boundary-based ranking loss function. This loss function compares the scores of true triples with those of negative sample triples, determining that the score of the true triple is lower than that of the negative sample triple, and the difference is not less than a preset interval parameter, thereby effectively maintaining the semantic consistency of the knowledge representation space. The above approach guides the vector representations of entities and relations to maintain the correct geometric relationships in the space, while improving the model's generalization ability when predicting unknown relations or inferring missing entities. Optionally, the TransE model also configures a weight update mechanism. This enables dynamic optimization of model parameters, allowing the positions of entity vectors and relation vectors in the representation space to continuously approach the true semantic relationships, improving the convergence speed and expression accuracy of the prediction model. Optionally, the TransE model normalizes the entity vectors after each round of parameter updates to constrain the norm of the entity vectors within a threshold range. This maintains numerical stability during training, ensures a balanced distribution of embedding vectors in space, and thus improves the overall robustness of the knowledge-enhanced prediction model.

[0088] The knowledge graph embedding model is configured with a negative sampling strategy. This strategy generates perturbation triples that disrupt the original correct entity-relation-entity pairings, randomly replacing the head or tail entity. This effectively expands the diversity of training samples, enabling the TransE model to better distinguish between valid and invalid semantic connections, thus improving the generalization ability and stability of the prediction model.

[0089] This disclosure employs a knowledge graph-based digital twin modeling method for manufacturing. This method fuses and constructs models from original multi-source data to obtain multi-source fused data and generate a digital twin architecture. Then, it performs ontological design and semantic modeling on the multi-source fused data to generate a manufacturing knowledge graph. Entity modeling is then performed on the manufacturing knowledge graph to obtain a set model, and reasoning is performed based on the manufacturing knowledge graph and historical data to obtain a knowledge-driven decision-making model. In this disclosure, the manufacturing knowledge graph is used to semantically express and reason about equipment, parts, processes, and their relationships in the manufacturing system. This not only improves data relevance and interpretability but also enables the digital twin architecture to possess knowledge-driven prediction and decision-making capabilities. Simultaneously, by directly applying intelligent decisions from the virtual space to the physical system through the decision model, it effectively improves system operating efficiency and enhances system robustness and adaptability, thus providing a highly robust and sustainably evolving digital twin modeling method for intelligent manufacturing, smart factories, and other fields. After obtaining the knowledge-driven decision model, this disclosure also embeds KGE (Knowledge-Based Engineering) into the decision model. KGE integrates manufacturing knowledge graph embedding with temporal prediction networks, enabling multi-source processing data to undergo speech enhancement and structural constraints before entering the prediction environment, thereby significantly improving the accuracy and robustness of predictions.

[0090] Optionally, the processor embeds KGE into the decision model to semantically enhance and structurally constrain the manufacturing knowledge graph before model prediction, including: the processor encodes and decodes multi-source fusion data through a knowledge-enhanced prediction model to obtain the original time series features; the processor performs model inference based on the knowledge-driven decision model and the original time series features to obtain the prediction results of the future manufacturing state of the manufacturing system.

[0091] Among them, the knowledge-enhanced prediction model not only provides feature supplementation during knowledge embedding, but also weights the temporal features of the multi-source fusion data through attention weights during the decoding stage, highlighting the process steps and equipment parameters relevant to the prediction task. The above process can be expressed by the following formula: .in, Features of the original time series Knowledge embedding features generated for knowledge-enhanced prediction models For the future The manufacturing status prediction results over time. This is the time-series prediction function for the knowledge-enhanced prediction model.

[0092] Optionally, the future manufacturing state prediction results include some or all of the predictions of key equipment health status or processing quality trends, and production cycle time analysis. The processor embeds KGE into the decision model to semantically enhance and structurally constrain the manufacturing knowledge graph before model prediction. This includes: the processor encoding and decoding multi-source fusion data through a knowledge-enhanced prediction model to obtain original time-series features; the processor performing model inference based on the knowledge-driven decision model and the original time-series features to obtain the future manufacturing state prediction results of the manufacturing system; and the processor inputting the future manufacturing state prediction results into the knowledge-driven decision model to generate optimization strategies. These optimization strategies include adjusting processing parameters, adjusting production sequence, or initiating preventative maintenance.

[0093] Combination Figure 9 As shown in the figure, this disclosure provides a knowledge graph-based digital twin modeling device 200 for manufacturing, including a virtual architecture construction module 201, a knowledge graph construction module 202, a geometry construction module 203, and a knowledge reasoning module 204. The virtual architecture construction module 201 is configured to perform data fusion and model building on original multi-source data to obtain multi-source fused data and generate a digital twin architecture. The knowledge graph construction module 202 is configured to perform ontological design and semantic modeling on the multi-source fused data to generate a manufacturing knowledge graph; wherein, the manufacturing knowledge graph is represented by triples, where triples include entity-relationship-entity / attribute. The geometry construction module 203 is configured to perform entity modeling on the manufacturing knowledge graph to obtain a knowledge-driven geometric model. The knowledge reasoning module 204 is configured to perform reasoning based on the manufacturing knowledge graph and historical data to obtain a knowledge-driven decision model.

[0094] The knowledge graph-based digital twin modeling device 200 for manufacturing provided in this disclosure generates a knowledge graph for semantic expression and reasoning of equipment, parts, processes, and their relationships within the manufacturing system. This not only improves data relevance and interpretability but also endows the digital twin architecture with knowledge-driven prediction and decision-making capabilities. Simultaneously, by directly applying intelligent decisions from the virtual space to the physical system through a decision model, it effectively improves system operating efficiency and enhances system robustness and adaptability. Therefore, it provides a highly robust and sustainably evolving digital twin modeling device for multiple fields such as intelligent manufacturing and smart factories.

[0095] Combination Figure 10As shown, this disclosure provides a knowledge graph-based digital twin modeling device 70 for manufacturing, including a processor 700 and a memory 701. Optionally, the device 70 may further include a communication interface 702 and a bus 703. The processor 700, communication interface 702, and memory 701 can communicate with each other via the bus 703. The communication interface 702 can be used for information transmission. The processor 700 can call logical instructions in the memory 701 to execute the knowledge graph-based digital twin modeling method for manufacturing described in the above embodiments.

[0096] Furthermore, the logic instructions in the aforementioned memory 701 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0097] The memory 701, as a computer-readable storage medium, can be used to store software programs and computer-executable programs, such as program instructions / modules corresponding to the methods in the embodiments of this disclosure. The processor 700 executes functional applications and data processing by running the program instructions / modules stored in the memory 701, thereby implementing the knowledge graph-based digital twin modeling method for manufacturing described above.

[0098] The memory 701 may include a program storage area and a data storage area. The program storage area may store the operating system and application programs required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory 701 may include high-speed random access memory and may also include non-volatile memory.

[0099] This disclosure provides an electronic device, including: a device body, and the aforementioned knowledge graph-based digital twin modeling device 200 (70) for manufacturing. The knowledge graph-based digital twin modeling device 200 (70) is installed on the device body. The installation relationship described herein is not limited to placement inside the device body, but also includes installation connections with other components of the electronic device, including but not limited to physical connections, electrical connections, or signal transmission connections. Those skilled in the art will understand that the knowledge graph-based digital twin modeling device 200 (70) for manufacturing can be adapted to feasible device bodies to achieve other feasible embodiments.

[0100] This disclosure provides a computer-readable storage medium storing computer-executable instructions configured to execute the above-described knowledge graph-based digital twin modeling method for manufacturing.

[0101] The technical solutions of this disclosure can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes one or more instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in this disclosure. The aforementioned storage medium can be a non-transitory storage medium, such as a USB flash drive, external hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, etc., and other media capable of storing program code.

[0102] The foregoing description and accompanying drawings fully illustrate embodiments of this disclosure to enable those skilled in the art to practice them. Other embodiments may include structural, logical, electrical, procedural, and other changes. The embodiments represent only possible variations. Individual components and functions are optional unless explicitly required, and the order of operation may vary. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. Moreover, the terminology used in this application is for describing embodiments only and is not intended to limit the claims. As used in the description of embodiments and claims, the singular forms “a,” “an,” and “the” are intended to equally include the plural forms unless the context clearly indicates otherwise. Similarly, the term “and / or” as used in this application means including one or more of the associated listed items and all possible combinations thereof. Additionally, when used in this application, the term "comprise" and its variations "comprises" and / or "comprising" refer to the presence of stated features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof. Without further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the process, method, or apparatus that includes said element. In this document, each embodiment may focus on the differences from other embodiments, and similar or identical parts between embodiments can be referred to mutually. For methods, products, etc., disclosed in the embodiments, if they correspond to the method section disclosed in the embodiments, the relevant parts can be referred to the description of the method section.

[0103] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of this disclosure. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

Claims

1. A knowledge graph-based digital twin modeling method for manufacturing, characterized in that, include: Data fusion and model building are performed on raw multi-source data to obtain multi-source fused data and generate a digital twin architecture; Ontological design and semantic modeling are performed on multi-source fusion data to generate a manufacturing knowledge graph; the manufacturing knowledge graph is represented by triples, which include entity-relationship-entity / attribute; Entity modeling is performed on the manufacturing knowledge graph to obtain a knowledge-driven geometric model; Based on manufacturing knowledge graphs and historical data, reasoning is used to obtain knowledge-driven decision-making models.

2. The method according to claim 1, characterized in that, Ontological design and semantic modeling of multi-source fused data are performed to generate a manufacturing knowledge graph, including: Entity identification, relation extraction, and attribute mapping are performed sequentially on the multi-source fused data to obtain triples represented by entity-relationship-entity / attribute; A manufacturing knowledge graph is constructed based on triples.

3. The method according to claim 1, characterized in that, Entity modeling is performed on the manufacturing knowledge graph to obtain a knowledge-driven geometric model, including: Based on the manufacturing knowledge graph, the system attribute information of the manufacturing system is obtained and the semantic rules of the manufacturing knowledge graph are constructed; among which, the attribute information includes component information, assembly relationships, and process constraints; Based on component information, establish the mapping relationship between key dimensions and knowledge parameters of geometric entities. Based on the semantic rules and mapping relationship of manufacturing knowledge graph, generate virtual object models corresponding to geometric entities and construct geometric models of virtual space based on virtual object models.

4. The method according to claim 3, characterized in that, Entity modeling of manufacturing knowledge graphs to obtain knowledge-driven geometric models also includes: Feature extraction and registration are performed on point cloud data to align it with the constructed geometric model, and model correction is performed on the virtual object model.

5. The method according to claim 1, characterized in that, Based on manufacturing knowledge graphs and historical data, reasoning is used to obtain knowledge-driven decision-making models, including: The production logic is obtained by retrieval and reasoning based on the manufacturing knowledge graph; the production logic includes resource constraints and process rules, and scheduling logic. Semantic parsing is performed on the production logic to build a reasoning logic library; the reasoning logic library is used to call the decision model. Key production data is mapped to semantic nodes of the manufacturing knowledge graph, thus integrating production data with the manufacturing knowledge graph; key production data includes equipment operating status and production progress, process parameters, and environmental information. A decision-making model is constructed in virtual space based on the semantic rules of the reasoning logic library and the knowledge graph.

6. The method according to claim 1, characterized in that, The raw multi-source data includes geometric measurement data, equipment status data, environmental monitoring data, and process record data; Data fusion and model building are performed on raw multi-source data to obtain multi-source fused data and generate a digital twin architecture, including: The original multi-source data is classified to obtain multi-source classified data; The multi-source classification data is cleaned, anomalies are removed, time sequence alignment is performed, and format is unified to obtain multi-source fused data. Construct a digital twin architecture; the digital twin architecture includes a physical layer, a data layer, and a knowledge-driven virtual layer. The physical layer is used to collect the operating status of equipment and processes, the data layer is used to aggregate, manage and process raw multi-source data, and the knowledge-driven virtual layer has semantic association and reasoning capabilities based on manufacturing knowledge graphs.

7. The method according to any one of claims 1 to 6, characterized in that, Also includes: After obtaining a knowledge-driven decision model by reasoning based on the manufacturing knowledge graph and historical data, the knowledge-enhanced prediction model KGE is embedded into the decision model to perform semantic enhancement and structural constraints on the manufacturing knowledge graph before model prediction.

8. A knowledge graph-based digital twin modeling device for manufacturing, characterized in that, include: The virtual architecture building module is configured to perform data fusion and model building on raw multi-source data to obtain multi-source fused data and generate a digital twin architecture; The knowledge graph construction module is configured to perform ontological design and semantic modeling on multi-source fusion data to generate a manufacturing knowledge graph; the manufacturing knowledge graph is represented by triples, which include entity-relationship-entity / attribute; The geometry building module is configured to perform entity modeling on the manufacturing knowledge graph to obtain a knowledge-driven geometric model; The knowledge reasoning module is configured to perform reasoning based on the manufactured knowledge graph and historical data to obtain a knowledge-driven decision model.

9. A knowledge graph-based digital twin modeling device for manufacturing, comprising a processor and a memory storing program instructions, characterized in that, The processor is configured to execute, when running the program instructions, the knowledge graph-based digital twin modeling method for manufacturing as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, include: Equipment body; The knowledge graph-based digital twin modeling device for manufacturing as described in claim 8 or 9 is installed on the device body.

Images