Industrial asset visualization method and system based on knowledge graph and digital twinning

Abstract

The application relates to the field of computers and digital twinning technology, and discloses an industrial asset visualization method and system based on a knowledge graph and digital twinning. The method comprises the following steps: fusing multi-source heterogeneous industrial asset data, constructing a knowledge graph based on an ontology model; establishing a high-fidelity digital twinning model, and realizing bidirectional mapping with the knowledge graph through a unique device identifier; synchronously rendering a three-dimensional geometric model and a semantic relationship network in a visualization interface, and supporting interactive linkage query and dynamic state display. The system comprises the following modules: multi-source data access, data standardization, knowledge graph construction, digital twinning modeling, bidirectional synchronization, visualization rendering, and interactive response. Through semantic and geometric deep fusion, the application realizes intelligent visualization of industrial assets in the whole-element and whole-life-cycle mode, and improves operation and maintenance decision efficiency and system scalability.

Description

Technical Field

[0001] This invention belongs to the field of computer and digital twin technology, specifically relating to a method and system for visualizing industrial assets based on knowledge graphs and digital twins. Background Technology

[0002] With the deepening of industrial digital transformation, the full lifecycle management of industrial assets places higher demands on visualization, intelligence, and collaboration. Current mainstream asset management systems mostly rely on relational database architectures, using equipment codes (such as KKS codes) to establish basic information associations, and introducing 3D models for geometric display in some scenarios. However, such systems are essentially still centered on static data tables, making it difficult to effectively depict the complex semantic relationships between assets in terms of physical connections, logical ownership, and operational dependencies.

[0003] Especially in large-scale process industries, the number of assets is huge, the types are diverse, and the topology changes dynamically. Traditional methods lack the ability to deeply integrate multi-source heterogeneous data, resulting in key information such as design drawings, equipment ledgers, real-time operating conditions and 3D models being in a state of long-term fragmentation, forming significant data silos.

[0004] Industrial asset visualization based on knowledge graphs and digital twins has become a research hotspot in recent years. This research aims to construct digital representations of assets that combine geometric form and semantic relationships through the fusion of semantic modeling and spatial mapping. The ideal technical approach should be able to automatically integrate multi-source data from PDMS, ERP, real-time databases, and document management systems, using a unified identifier as an anchor point to dynamically construct asset entities and their associated networks, and achieve bidirectional interaction between the knowledge graph and the 3D model at the visualization level.

[0005] However, existing attempts are generally limited by problems such as reliance on static models, manual relationship maintenance, and weak semantic expression capabilities, which cannot support real-time simulation and intuitive presentation of asset state evolution, fault propagation paths, or the impact of security risks.

[0006] Existing technologies generally suffer from the following problems: heterogeneous data formats and inconsistent standards from multiple sources, lacking a unified encoding mapping and semantic alignment mechanism, resulting in low data integration efficiency; the modeling of relationships between assets relies heavily on manual input, which is slow to update and prone to errors, making it difficult to reflect dynamic topological changes in actual operation; 3D visualization only displays geometric shapes and does not establish dynamic binding with the asset's operating parameters, logical systems, or security attributes; more importantly, there is a lack of an architecture that deeply couples the semantic reasoning capabilities of knowledge graphs with the spatial representation capabilities of digital twins, making it impossible for operations and maintenance personnel to simultaneously understand "where the assets are" and "what the assets mean" in a unified interface.

[0007] The aforementioned problems are particularly prominent in high-risk industrial scenarios. Once equipment malfunctions, the system struggles to quickly trace the affected links and assess the scope of risks, severely limiting the response speed and decision-making accuracy of safety management. Therefore, there is an urgent need for an intelligent industrial asset management solution that can achieve multi-source data fusion, dynamic relationship modeling, and semantic-spatial dual visualization. Summary of the Invention

[0008] This invention provides an industrial asset visualization method and system based on knowledge graphs and digital twins, aiming to solve the technical problems existing in the current industrial asset management system, such as data silos, weak asset relationship expression capabilities, lack of dynamic semantic associations, and visualization limited to the geometric level and unable to reflect semantic relationships.

[0009] This invention constructs a unified data fusion framework to structurally model multi-source heterogeneous industrial asset data, and uses knowledge graph technology to dynamically semantically model the logical, physical, functional, and operational relationships between assets. At the same time, it combines a high-fidelity digital twin model to achieve real-time linkage and bidirectional mapping between asset status, operating parameters, drawings and documents, and three-dimensional geometric models. Finally, the geometric shape and semantic relationship network of the assets are synchronously presented in a visualization interface, thereby providing full-element, full-lifecycle, and full-dimensional visualization support for industrial asset management.

[0010] This invention provides a method for visualizing industrial assets based on knowledge graphs and digital twins, including: Obtain raw industrial asset data from multiple data sources. The raw data of the industrial assets is standardized, cleaned, and entity is identified to extract equipment entities, attribute information, relation triples, and spatiotemporal state parameters. An industrial asset knowledge graph is constructed based on a predefined ontology model. The ontology model defines the hierarchical structure and constraint rules of equipment classes, attribute classes, relationship classes and state classes. The equipment entities, attribute information and relationship triples are injected into the ontology model to generate a knowledge graph instance with semantic consistency. Construct a digital twin model that corresponds one-to-one with the physical factory; Establish a two-way data channel between the knowledge graph and the digital twin model. When the device status in the knowledge graph is updated, the visual status change of the corresponding geometric object in the digital twin model is triggered by the device identifier. Conversely, when the user selects a geometric object in the digital twin model, the system automatically queries the knowledge graph and displays the complete attributes of the device, associated devices, and historical status sequence. The digital twin model and the knowledge graph relationship network are rendered synchronously in the visualization interface. The knowledge graph relationship network is overlaid on the three-dimensional scene in a force-directed graph layout or presented as an independent view. Nodes represent device entities, edges represent semantic relationships, node colors and sizes are dynamically adjusted according to real-time status parameters, and edge thickness and type are distinguished according to relationship strength and category. In response to user interaction, when a user clicks on any node in the knowledge graph, the system highlights the corresponding geometric object in the digital twin model and automatically focuses on the spatial area where the geometric object is located, while loading its associated documents, historical alarm records and maintenance work order list.

[0011] Preferably, the raw industrial asset data undergoes standardized cleaning and entity identification, including: A hybrid named entity recognition model combining rule-based and deep learning is used to extract device names, models, and specifications from unstructured text data. Perform field alignment and unit standardization on structured tabular data; The component tree structure of the 3D model file is parsed, and the unique code and spatial bounding box information of each component are extracted. Sliding window aggregation is performed on real-time sensor data streams to generate timestamped state snapshots.

[0012] Preferably, a bidirectional data channel is established between the knowledge graph and the digital twin model, including: Embed a device uniform identifier in the metadata of each geometric object in the digital twin model; The identifier is stored in each device entity node of the knowledge graph; An intermediate message bus is constructed to subscribe to real-time status data streams. When new status data is received, its device identifier is parsed, the dynamic attribute values ​​of the corresponding node in the knowledge graph are updated, and rendering instructions are sent to the visualization engine to modify the color, transparency, or additional status icons of the corresponding objects in the 3D model.

[0013] Preferably, a device uniform identifier is embedded in the metadata of each geometric object in the digital twin model, including: High-precision 3D geometric models are generated by laser scanning point cloud reconstruction or CAD model conversion. Bind a device unique identifier to each geometric object that is exactly the same as the device entity URI in the knowledge graph; Align the model coordinate system with the actual geographical coordinate system of the factory to ensure spatial consistency.

[0014] Preferably, constructing an intermediate message bus includes: Subscribe to real-time status data streams from sensor networks and monitoring systems; The device uniform identifier mapping table is stored in an in-memory database, supporting millisecond-level query response; When a state update event is triggered, the knowledge graph node attributes are updated synchronously and rendering instructions are broadcast to the visualization engine to ensure that the synchronization delay between the 3D model and the knowledge graph is less than 500 milliseconds.

[0015] Preferably, the digital twin model and the knowledge graph relationship network are rendered simultaneously in a visual interface, including: The WebGL engine is used to render the 3D scene, and D3.js is used to implement the 2D force-guided layout of the knowledge graph. Spatial and temporal alignment of dual views is achieved by sharing a coordinate system and a unified time axis; Based on user-defined relationship types, device categories, or status thresholds, the knowledge graph topology is dynamically filtered and reconstructed, and the visibility and highlight status of the 3D model are updated synchronously.

[0016] Preferably, responding to user interaction operations includes: When a user clicks on any node in the knowledge graph, the corresponding geometric object in the digital twin model is highlighted and automatically focused on the spatial region where the geometric object is located; Call the document management interface to load associated PDF drawings, call the work order system interface to retrieve the three most recent maintenance records, and display them in a structured table format in the sidebar; In a 3D scene, mark the upstream and downstream related devices of the device. The upstream and downstream related devices include adjacent devices that are physically connected, functionally dependent, or operationally related.

[0017] Preferably, sliding window aggregation of real-time sensor data streams includes: Set the window length to 5 seconds and collect raw data at a sampling frequency of 10 times per second; Calculate the mean, maximum, minimum and standard deviation within the window, and generate a state snapshot with a uniform timestamp; Apply the three sigma criterion to remove outliers and output a standardized state snapshot.

[0018] Preferably, the device entity, attribute information, and relation triples are injected into the ontology model, including: Verify whether the input triples satisfy the ontology axiom constraints. If conflicting triples are detected, trigger a manual verification process or automatically resolve them according to preset priority rules. The validated triples are stored in the graph database in RDF format. Each node contains a unique URI identifier, category label, attribute dictionary and state timestamp, and each edge contains relation type, directionality and confidence weight.

[0019] This invention also provides an industrial asset visualization system based on knowledge graphs and digital twins, including: The multi-source data access module is used to acquire raw industrial asset data from equipment ledger databases, real-time monitoring systems, historical work order records, 3D design drawings, equipment manual documents, and sensor networks. The data standardization and entity recognition module is used to standardize and clean the raw data of the industrial assets, and extract equipment entities, attribute information, relation triples and spatiotemporal state parameters. The knowledge graph construction module is used to inject the device entities, attribute information and relation triples into the predefined ontology model to generate an industrial asset knowledge graph. The digital twin model building module is used to construct a high-precision 3D geometric model corresponding to the physical factory and bind a unique device identifier to each geometric object; The bidirectional mapping and synchronization module is used to establish a bidirectional data channel between the knowledge graph and the digital twin model, enabling real-time linkage and interactive response for state updates; The visualization rendering module is used to synchronously render the digital twin model and the knowledge graph relationship network in the user interface, and supports interactive focusing and information loading based on user interaction. The interactive response processing module is used to respond to user clicks, drags, and filtering operations on knowledge graph nodes or 3D model objects, and to execute corresponding data queries, view switching, and status highlighting commands.

[0020] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention fundamentally solves the problems of data silos and static relationships in traditional asset management systems by deeply integrating the semantic modeling capabilities of knowledge graphs with the geometric simulation capabilities of digital twins. Multi-source heterogeneous data, after standardization processing, is uniformly injected into the ontology model to form an industrial asset knowledge graph with logical consistency and dynamic updating capabilities, avoiding the lag and error rate of manually maintaining relationship tables.

[0021] 2. The digital twin model and knowledge graph establish a two-way mapping through a unique device identifier, realizing real-time linkage between geometric objects and semantic entities. This allows changes in asset status to be reflected instantly on the 3D visualization interface, while user operations on the 3D model can directly drive the query and display of semantic information.

[0022] 3. At the visualization level, not only is a high-fidelity spatial geometric representation retained, but an interactive, filterable, and focusable semantic relationship network is also superimposed, enabling maintenance personnel to intuitively understand the complex dependencies between devices, quickly locate fault propagation paths, and improve the intelligence level and decision-making efficiency of asset management.

[0023] 4. The introduction of the ontology model ensures the standardization and scalability of data integration, providing a standardized framework for subsequent access to new types of data sources or additions of device categories, significantly enhancing the system's adaptability and lifecycle value. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the overall technical solution architecture of the present invention; Figure 2 This is a schematic diagram of the core principle framework of the bidirectional mapping and synchronization mechanism between knowledge graph and digital twin in this invention; Figure 3 This is a logical flowchart of the multi-source industrial asset data standardization cleaning and entity recognition in this invention; Figure 4 This is a logical flowchart of the construction of an industrial asset knowledge graph based on an ontology model and the guarantee of semantic consistency in this invention. Figure 5 This is a schematic diagram of the multi-level interaction relationship and data flow of the synchronous rendering and interactive response of the digital twin model and the knowledge graph relationship network in this invention; Figure 6 This is a schematic diagram of the multi-level interactive relationship and data flow of three-dimensional geometric objects and semantic relationship network linkage focusing and information loading driven by user interaction in this invention. Detailed Implementation

[0025] refer to Figures 1 to 6 This invention provides a method and system for visualizing industrial assets based on knowledge graphs and digital twins. The method constructs a unified data fusion framework to structurally model multi-source heterogeneous industrial asset data. It then utilizes knowledge graph technology to dynamically semantically model the logical, physical, functional, and operational relationships between assets. Simultaneously, it combines a high-fidelity digital twin model to achieve real-time linkage and bidirectional mapping between asset status, operating parameters, drawings, and 3D geometric models. Finally, the geometric form and semantic relationship network of the assets are synchronously presented in the visualization interface.

[0026] The method includes the following steps: S1, acquire raw industrial asset data from multiple data sources; S2, perform standardized cleaning and entity recognition on the raw data of the industrial assets; S3, constructs an industrial asset knowledge graph based on a predefined ontology model; S4, constructing a digital twin model that corresponds one-to-one with the physical factory; S5 establishes a two-way data channel between knowledge graphs and digital twin models; S6 synchronously renders the digital twin model and the knowledge graph relationship network in a visual interface; S7 responds to user interaction.

[0027] In step S1, raw industrial asset data is acquired from multiple data sources, including an equipment ledger database, a real-time monitoring system, historical work order records, 3D design drawings, equipment manuals, and sensor networks. The equipment ledger database stores basic equipment information in a relational table format, with fields including equipment code, equipment name, system affiliation, installation location, commissioning time, manufacturer, model specifications, and maintenance cycle. The real-time monitoring system continuously pushes equipment operating parameters, including current temperature, pressure, vibration level, current, voltage, start / stop status, and valve opening, via industrial communication protocols such as Modbus TCP or OPC UA.

[0028] Historical work order records are stored in structured log format, including the time, personnel involved, fault description, handling measures, and list of replaced parts for each maintenance, inspection, and replacement operation. 3D design drawings are in IFC or OBJ format, containing a complete factory geometry and equipment component tree, with each component having a unique code and bounding box information.

[0029] The equipment manual is an unstructured PDF or DOC document containing technical parameters, wiring diagrams, operation manuals, and safety specifications. The sensor network consists of wireless or wired sensor nodes deployed in the field, collecting environmental and equipment status data at a fixed frequency, including geographic coordinates, temperature and humidity, acceleration, and acoustic emission signals. All data sources undergo protocol adaptation and format conversion through a unified access gateway, ensuring that raw data enters subsequent processing flows in a standardized message format.

[0030] In step S2, the raw industrial asset data is standardized, cleaned, and entity-identified to extract equipment entities, attribute information, relational triples, and spatiotemporal state parameters. Equipment entities include core industrial asset units such as pumps, valves, pipes, motors, and control cabinets. Attribute information includes static features such as equipment model, installation location, commissioning time, and maintenance cycle. Relational triples include semantic associations such as "belongs to," "connected to," "drives," "controls," and "located in the same subsystem." Spatiotemporal state parameters include dynamic indicators such as current temperature, pressure, vibration value, start / stop status, and geographical coordinates.

[0031] Specifically, this includes processing unstructured text data using a hybrid named entity recognition model that combines rule-based and deep learning. The model first matches typical formats of equipment models and specifications using regular expressions, and then models the contextual semantics using a pre-trained BERT-BiLSTM-CRF neural network to accurately extract equipment names, technical parameters, and their units.

[0032] Perform field alignment and unit standardization on structured tabular data. For example, map the field representing "installation location" from different databases to "installation location" and standardize the pressure unit to megapascal and the temperature unit to Celsius.

[0033] The component tree structure of the 3D model file is parsed, the hierarchical nodes of the IFC or OBJ file are traversed, and the globally unique identifier, parent node reference, geometry type and spatial bounding box coordinate range of each component are extracted.

[0034] A sliding window aggregation method is used to aggregate real-time sensor data streams, with a window length of 5 seconds. Raw data is collected at a sampling frequency of 10 times per second. Within the window, the mean, maximum, minimum, and standard deviation are calculated to generate a state snapshot with a uniform timestamp. Three Sigma criteria are applied to remove outliers. All cleaned data is encapsulated into structured JSON objects, containing entity identifiers, attribute key-value pairs, relationship source and target, and state time series.

[0035] In step S3, an industrial asset knowledge graph is constructed based on a predefined ontology model. The ontology model is constructed using a descriptive logic language, defining the hierarchical structure and constraint rules for equipment classes, attribute classes, relationship classes, and state classes. The top-level equipment class is Device, which includes subclasses for mechanical equipment, electrical equipment, and control equipment. Mechanical equipment includes pumps, valves, and pipelines; electrical equipment includes motors, transformers, and distribution cabinets; and control equipment includes PLCs, DCS controllers, and human-machine interfaces.

[0036] The relationship classes include physical connection relationships, functional dependency relationships, and operation and maintenance association relationships. The attribute classes are divided into static attributes and dynamic attributes. Static attributes cover inherent device parameters such as model, manufacturer, and rated power, while dynamic attributes are associated with real-time data interface addresses such as Modbus register addresses or OPC UA node paths. The device entities, attribute information, and relationship triples extracted in step S2 are injected into the ontology model to generate a semantically consistent knowledge graph instance.

[0037] The injection process adheres to ontology axioms, such as "pump" must be "connected to" a pipeline, and "motor" must "drive" the pump. If a triple violating these axioms is detected, such as "valve drives motor," the system triggers a manual verification process or automatically resolves the violation based on preset priority rules. The knowledge graph is stored in a graph database as RDF triples. Each node contains a unique URI identifier, category label, attribute dictionary, and state timestamp. Each edge contains relation type, directionality, and confidence weight.

[0038] In step S4, a digital twin model corresponding one-to-one with the physical factory is constructed. The digital twin model consists of a high-precision three-dimensional geometric model, generated through laser scanning point cloud reconstruction or CAD model conversion, with a geometric accuracy error of less than 5 millimeters. Each geometric object corresponds to a physical device or component, and its metadata embeds a unique device identifier, which is completely consistent with the URI of the corresponding device entity in the knowledge graph.

[0039] The initial values ​​for the material, color, and transparency of geometric objects are preset according to the equipment type; for example, the pump body is made of gray metal, the pipe is blue, and the cable is black. The model's coordinate system is aligned with the actual geographical coordinate system of the factory to ensure spatial consistency. The digital twin model is organized with a hierarchical LOD (Level of Detail) structure, supporting seamless switching between long-distance overview and close-up detailed viewing.

[0040] In step S5, a bidirectional data channel is established between the knowledge graph and the digital twin model. This bidirectional data channel uses a unified device identifier to create an index mapping table, which is stored in an in-memory database Redis, supporting millisecond-level query responses. An intermediate message bus is constructed, using Kafka as the message middleware, to subscribe to real-time status data streams.

[0041] When new status data is received, the message bus parses its device identifier, queries the corresponding node in the knowledge graph, and updates its dynamic attribute values, including current temperature, pressure, and vibration values. Simultaneously, it sends rendering commands to the visualization engine to modify the color, transparency, or add status icons to the corresponding geometric objects in the 3D model. For example, when the pump's vibration value exceeds a threshold, the geometric object changes from gray to red and flashes; when a valve is closed, its handwheel displays a locked icon.

[0042] Conversely, when a user selects a geometric object in the digital twin model, the system queries the knowledge graph using its embedded device identifier to obtain the device's complete attributes, a list of associated devices, and a historical state sequence, which is then displayed in the sidebar. The bidirectional channel ensures that the synchronization latency for state updates is less than 500 milliseconds.

[0043] In step S6, the digital twin model and the knowledge graph relationship network are simultaneously rendered in the visualization interface. The digital twin model uses a WebGL engine such as Three.js to render a 3D scene, supporting realistic effects such as lighting, shadows, and reflections. The knowledge graph relationship network uses D3.js to implement a 2D force-guided layout, where nodes represent device entities and edges represent semantic relationships. The relationship network can be overlaid on the 3D scene as a semi-transparent overlay, or presented side-by-side as independent views.

[0044] Node colors are dynamically adjusted based on real-time status parameters; for example, higher temperatures result in redder colors, and greater vibration values ​​lead to larger node sizes. Edge thickness is determined by relationship strength; for example, edges between frequently co-occurring device pairs are thicker. Edge types are distinguished by relationship category: solid lines for physical connections, dashed lines for functional dependencies, and dotted lines for operational associations. The system provides a hierarchical filtering mechanism, allowing users to dynamically filter by relationship type, device category, and status threshold.

[0045] For example, only device nodes with "physical connection" relationships and temperatures greater than 80 degrees Celsius are displayed. When the filtering criteria change, the system reconstructs the map topology in real time and updates the visibility and highlight status of the 3D model synchronously, hiding irrelevant devices and highlighting the filtering results.

[0046] In step S7, the system responds to user interaction. When the user clicks on any node in the knowledge graph, the system highlights the corresponding geometric object in the digital twin model, sets its color to bright yellow, and automatically focuses on the spatial area where the geometric object is located, adjusting the camera angle to center it. Simultaneously, the system calls the document management interface to load associated PDF drawings, calls the work order system interface to retrieve the three most recent maintenance records, and displays the equipment model, installation location, commissioning time, current status, associated equipment list, historical alarm records, and maintenance work order summary in a structured table format in the sidebar.

[0047] Furthermore, the system marks upstream and downstream related devices of a device in the 3D scene. For example, for a pump, it highlights its upstream inlet pipe, downstream outlet pipe, drive motor, and control valve. Users can also directly click on geometric objects in the 3D model to trigger the same information loading and graph focusing process. Drag and drop operations allow users to rearrange knowledge graph nodes, and the system records and persistently stores user-defined layouts. All interactive events are distributed through a unified event bus, ensuring loose coupling and high responsiveness between modules.

[0048] The system includes a multi-source data access module, a data standardization and entity recognition module, a knowledge graph construction module, a digital twin model construction module, a bidirectional mapping and synchronization module, a visualization rendering module, and an interactive response processing module.

[0049] The multi-source data access module is equipped with various protocol adapters, including an ODBC / JDBC adapter for connecting to relational databases, an OPCUA client for interfacing with industrial control systems, an FTP / HTTP client for downloading drawings and documents, and an MQTT subscriber for receiving sensor data streams. All access data is converted into a unified internal message format after protocol parsing.

[0050] The data standardization and entity recognition module includes an unstructured text parsing unit, a structured data alignment unit, a 3D model parsing unit, and a real-time data aggregation unit. The unstructured text parsing unit loads a pre-trained hybrid named entity recognition model to perform word segmentation, part-of-speech tagging, named entity recognition, and relation extraction on the input text. The structured data alignment unit maintains a field mapping table and a unit conversion table, performing field renaming and numerical conversion. The 3D model parsing unit calls the IFC parsing library or OBJ reader, recursively traversing the model tree to extract component metadata. The real-time data aggregation unit maintains a sliding window buffer, performing time alignment, statistical aggregation, and anomaly detection.

[0051] The knowledge graph construction module loads a predefined ontology model, verifies the logical consistency of the input triples, and performs graph injection and conflict resolution. The graph is stored in the Neo4j graph database and supports SPARQL queries and graph algorithm analysis.

[0052] The digital twin model building module integrates a point cloud processing and CAD conversion toolchain to generate a 3D model with metadata binding. The model is exported in glTF format, with device identifiers embedded in the extras field of each node.

[0053] The bidirectional mapping and synchronization module maintains a device identifier mapping table in memory, listens for knowledge graph change events and sensor data streams, and triggers 3D model state updates. The message bus adopts a publish-subscribe pattern to ensure low-latency synchronization.

[0054] The visualization rendering module initializes the WebGL rendering context and D3 force-directed graph layout, registers user interaction event listeners, and performs spatial and temporal alignment between the two views. A shared coordinate system ensures that the spatial positions in the 3D model and the logical positions of the graph nodes are consistent with the user's perception.

[0055] The interaction response processing module parses user clicks, drags, and filtering operations, and executes instructions such as device focus, information loading, and association tagging. All external system calls are completed through RESTful APIs or gRPC interfaces, ensuring module decoupling.

[0056] This embodiment, through the above-described method and system, achieves full-element, full-lifecycle, and full-dimensional visualization of industrial assets, solving the technical problems of data silos, static relationships, lack of dynamic semantic associations, and visualization limited to the geometric level.

[0057] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0058] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for visualizing industrial assets based on knowledge graphs and digital twins, characterized in that, include: Obtain raw industrial asset data from multiple data sources. The raw data of the industrial assets is standardized, cleaned, and entity is identified to extract equipment entities, attribute information, relation triples, and spatiotemporal state parameters. An industrial asset knowledge graph is constructed based on a predefined ontology model. The ontology model defines the hierarchical structure and constraint rules of equipment classes, attribute classes, relationship classes and state classes. The equipment entities, attribute information and relationship triples are injected into the ontology model to generate a knowledge graph instance with semantic consistency. Construct a digital twin model that corresponds one-to-one with the physical factory; Establish a two-way data channel between the knowledge graph and the digital twin model. When the device status in the knowledge graph is updated, the visual status change of the corresponding geometric object in the digital twin model is triggered by the device identifier. Conversely, when the user selects a geometric object in the digital twin model, the system automatically queries the knowledge graph and displays the complete attributes of the device, associated devices, and historical status sequence. The digital twin model and the knowledge graph relationship network are rendered synchronously in the visualization interface. The knowledge graph relationship network is overlaid on the three-dimensional scene in a force-directed graph layout or presented as an independent view. Nodes represent device entities, edges represent semantic relationships, node colors and sizes are dynamically adjusted according to real-time status parameters, and edge thickness and type are distinguished according to relationship strength and category. In response to user interaction, when a user clicks on any node in the knowledge graph, the system highlights the corresponding geometric object in the digital twin model and automatically focuses on the spatial area where the geometric object is located, while loading its associated documents, historical alarm records and maintenance work order list.

2. The industrial asset visualization method based on knowledge graphs and digital twins according to claim 1, characterized in that, The raw data of the industrial assets is standardized, cleaned, and entity-identified, including: A hybrid named entity recognition model combining rule-based and deep learning is used to extract device names, models, and specifications from unstructured text data. Perform field alignment and unit standardization on structured tabular data; The component tree structure of the 3D model file is parsed, and the unique code and spatial bounding box information of each component are extracted. Sliding window aggregation is performed on real-time sensor data streams to generate timestamped state snapshots.

3. The industrial asset visualization method based on knowledge graphs and digital twins according to claim 2, characterized in that, Establish a two-way data channel between knowledge graphs and digital twin models, including: Embed a device uniform identifier in the metadata of each geometric object in the digital twin model; The identifier is stored in each device entity node of the knowledge graph; An intermediate message bus is constructed to subscribe to real-time status data streams. When new status data is received, its device identifier is parsed, the dynamic attribute values ​​of the corresponding node in the knowledge graph are updated, and rendering instructions are sent to the visualization engine to modify the color, transparency, or additional status icons of the corresponding objects in the 3D model.

4. The industrial asset visualization method based on knowledge graphs and digital twins according to claim 3, characterized in that, Embed device uniform identifiers in the metadata of each geometric object in the digital twin model, including: High-precision 3D geometric models are generated by laser scanning point cloud reconstruction or CAD model conversion. Bind a device unique identifier to each geometric object that is exactly the same as the device entity URI in the knowledge graph; Align the model coordinate system with the actual geographical coordinate system of the factory to ensure spatial consistency.

5. The industrial asset visualization method based on knowledge graphs and digital twins according to claim 4, characterized in that, Constructing an intermediate message bus includes: Subscribe to real-time status data streams from sensor networks and monitoring systems; The device uniform identifier mapping table is stored in an in-memory database, supporting millisecond-level query response; When a state update event is triggered, the knowledge graph node attributes are updated synchronously and rendering instructions are broadcast to the visualization engine to ensure that the synchronization delay between the 3D model and the knowledge graph is less than 500 milliseconds.

6. The industrial asset visualization method based on knowledge graphs and digital twins according to claim 5, characterized in that, The digital twin model and the knowledge graph relationship network are rendered synchronously in a visual interface, including: The WebGL engine is used to render the 3D scene, and D3.js is used to implement the 2D force-guided layout of the knowledge graph. Spatial and temporal alignment of dual views is achieved by sharing a coordinate system and a unified time axis; Based on user-defined relationship types, device categories, or status thresholds, the knowledge graph topology is dynamically filtered and reconstructed, and the visibility and highlight status of the 3D model are updated synchronously.

7. The industrial asset visualization method based on knowledge graphs and digital twins according to claim 6, characterized in that, Responding to user interaction actions, including: When a user clicks on any node in the knowledge graph, the corresponding geometric object in the digital twin model is highlighted and automatically focused on the spatial region where the geometric object is located; Call the document management interface to load associated PDF drawings, call the work order system interface to retrieve the three most recent maintenance records, and display them in a structured table format in the sidebar; In a 3D scene, mark the upstream and downstream related devices of the device. The upstream and downstream related devices include adjacent devices that are physically connected, functionally dependent, or operationally related.

8. The industrial asset visualization method based on knowledge graphs and digital twins according to claim 7, characterized in that, Sliding window aggregation of real-time sensor data streams includes: Set the window length to 5 seconds and collect raw data at a sampling frequency of 10 times per second; Calculate the mean, maximum, minimum and standard deviation within the window, and generate a state snapshot with a uniform timestamp; Apply the three sigma criterion to remove outliers and output a standardized state snapshot.

9. The industrial asset visualization method based on knowledge graphs and digital twins according to claim 8, characterized in that, Injecting the device entity, attribute information, and relation triples into the ontology model includes: Verify whether the input triples satisfy the ontology axiom constraints. If conflicting triples are detected, trigger a manual verification process or automatically resolve them according to preset priority rules. The validated triples are stored in the graph database in RDF format. Each node contains a unique URI identifier, category label, attribute dictionary and state timestamp, and each edge contains relation type, directionality and confidence weight.

10. An industrial asset visualization system based on knowledge graphs and digital twins, characterized in that: include: The multi-source data access module is used to acquire raw industrial asset data from equipment ledger databases, real-time monitoring systems, historical work order records, 3D design drawings, equipment manual documents, and sensor networks. The data standardization and entity recognition module is used to standardize and clean the raw data of the industrial assets, and extract equipment entities, attribute information, relation triples and spatiotemporal state parameters. The knowledge graph construction module is used to inject the device entities, attribute information and relation triples into the predefined ontology model to generate an industrial asset knowledge graph. The digital twin model building module is used to construct a high-precision 3D geometric model corresponding to the physical factory and bind a unique device identifier to each geometric object; The bidirectional mapping and synchronization module is used to establish a bidirectional data channel between the knowledge graph and the digital twin model, enabling real-time linkage and interactive response for state updates; The visualization rendering module is used to synchronously render the digital twin model and the knowledge graph relationship network in the user interface, and supports interactive focusing and information loading based on user interaction. The interactive response processing module is used to respond to user clicks, drags, and filtering operations on knowledge graph nodes or 3D model objects, and to execute corresponding data queries, view switching, and status highlighting commands.

Images