Method for generating a digital twin of a system

An AI-driven knowledge graph solution structures and automates the creation of a digital twin by defining hierarchical and relational plant information, addressing the challenges of diverse and unstructured data sources, enabling efficient retrieval and control of plant components.

WO2026233496A1PCT designated stage Publication Date: 2026-11-12BAYERISCHE MOTOREN WERKE AG
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/DE2026/100500
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-05-09
Filing Date
2026-04-17
Publication Date
2026-11-12

AI Technical Summary

Technical Problem

Existing methods struggle to efficiently structure and automate the creation of a digital twin of a plant due to diverse, unstructured, and error-prone plant information from various sources, lacking standardized documentation and unclear relationships between components.

Method used

Employing an artificial intelligence algorithm, particularly generative AI, to generate a knowledge graph that defines the hierarchical structure and relationships between assemblies and components within a plant, using plant information from diverse sources and formats, and corrects errors through user input and external data verification.

Benefits of technology

Enables structured and automated assignment of plant information to specific assemblies and components, facilitating quick retrieval and control of plant components, reducing user workload and enhancing data usability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure DE2026100500_12112026_PF_FP_ABST
    Figure DE2026100500_12112026_PF_FP_ABST
Patent Text Reader

Abstract

The invention relates to a method for generating a digital twin of a system, comprising a plurality of assemblies having a plurality of components, wherein system information that describes the system is provided, in particular from different sources (3) and / or in different formats; and the digital twin is generated, by means of at least one artificial intelligence algorithm, as a knowledge graph (2), the knowledge graph describing hierarchical information, which describes the hierarchical structure of the assemblies within the system, and relationship information, which describes the relationship of the assemblies and / or components within the system.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] 25-0645 DE

[0002] Method for creating a digital twin of a plant

[0003] The invention relates to a method for generating a digital twin of a plant comprising several assemblies with multiple components, wherein plant-descriptive information, in particular from different sources and / or in different formats, is provided.

[0004] It is known from the prior art that plant information for a specific plant, for example, for the manufacture of motor vehicles, is provided from various sources and / or in different formats and can be stored in a database. Due to the complexity of such plants, where individual assemblies and components are usually supplied by different manufacturers, the plant information is correspondingly diverse. Structuring or digitizing the plant information is therefore only possible to a limited extent, as the provided documentation is mostly not standardized.

[0005] It is also possible that documentation or provided plant information contains errors, or that errors may occur during the transfer to the plant's digital twin due to the high manual effort involved. In addition to the fact that the plant information is very complex and encompasses various types of information, such as circuit diagrams, tables, texts, three-dimensional models, and the like, the relationships between the information within the plant information itself are unknown. In other words, it is not known which information pertains to which individual components and assemblies.

[0006] Furthermore, since it is not known which information from the overall plant information should be assigned to which assembly or component of the plant, a corresponding assignment, for example spatially or hierarchically, with regard to the executed process and the like, is not possible automatically. This results in the plant information being stored for the entire plant, making it difficult and time-consuming to retrieve information for individual assemblies or components. 25-0645 DE

[0007] The invention is based on the objective of providing an improved method for generating a digital twin of a plant, in which the provided plant information can be used in an improved automated manner.

[0008] The problem is solved by a method according to claim 1. The dependent claims relate to possible embodiments.

[0009] As described, the invention relates to a method for generating a digital twin of a plant. For the purposes of this description, a "digital twin" is understood to be a digital representation of a real plant. The digital twin contains, for example, data about the plant provided based on the plant information, so that this data can be accessed using the digital twin, for example, by a user. The plant comprises several assemblies with multiple components, for example, subsystems or subgroups.

[0010] To generate the digital twin of the plant, the process is provided with plant information that describes the plant. This plant information is provided from various sources and / or in different formats. "Plant information" can be understood as any information relating to the plant, i.e., at least one assembly and / or at least one component of the plant. Examples of plant information include three-dimensional models, parts lists, manuals, maintenance instructions, circuit diagrams, data connections, and the like. The plant information is typically provided to the process in an unstructured manner. This means that the process receives the plant information as a "pool" or the entirety of the provided data, without any assignment of the individual data points to assemblies or components.

[0011] The proposed method is based on the understanding that, using at least one artificial intelligence algorithm, in particular generative artificial intelligence, the digital twin is generated as a knowledge graph that describes hierarchy information relating to the hierarchical structure of the assemblies within the plant and relationship information relating to the relationships between the assemblies and / or components within the plant. This is referred to as a “knowledge graph” 25-0645 DE

[0012] Within the scope of this description, a generally known "knowledge graph" is understood in particular.

[0013] In other words, artificial intelligence is used to generate a knowledge graph that describes the hierarchical structure and relationships of the assemblies and / or components through hierarchy and relationship information. The generated hierarchy information describes the hierarchical structure of the assemblies within the plant, for example, which assemblies or components are assigned to which hierarchical levels. For instance, the entire plant itself could be the highest hierarchical level, specifically the topmost node in the knowledge graph. Below this, a machine level could be formed as a hierarchical level, containing various machines assigned to the plant, such as individual handling devices, processing devices, conveyors, and the like.Each such machine can in turn be subdivided into its assemblies at a further hierarchical level, which in turn can be subdivided into components at a lower hierarchical level, and these in turn into standard parts.

[0014] The hierarchical structure of the assemblies within the system allows for improved assignment of system information, or parts thereof, based on the hierarchy information. This assignment can be made to the specific assembly or component affected or described by that piece of information. In particular, system information can be precisely linked to the hierarchical level for which it is relevant. For example, if system information pertains to a specific component, it can be assigned to that component within the hierarchical structure. Similarly, if system information pertains to a three-dimensional model of a particular assembly, it can be assigned to that assembly at the relevant hierarchical level.

[0015] Furthermore, the knowledge graph describes the relationship information, or rather, the knowledge graph includes such relationship information that describes the relationships between the assemblies and / or components within the system. This relationship information describes, for example, the structural and / or functional relationships between the assemblies and / or components of the system. Such a 25-0645 DE

[0016] A relationship can be, for example, a mechanical relationship, an electrical relationship, and / or a data relationship. The relationship information indicates, for example, which components are electrically connected, interact mechanically, or exchange data.

[0017] In other words, the plant information can still be provided in any format and / or from various sources. Using the described algorithm, the knowledge graph is generated, which describes the hierarchy and relationship information that defines the hierarchical structure of the assemblies and the relationships between the assemblies and / or components. The algorithm can extract relevant information from the different sources and formats and, based on this, determine the hierarchical structure of the plant as well as the relationships between the assemblies and / or components. The information thus obtained, particularly the relationship and hierarchy information, is then output as a knowledge graph, or the knowledge graph is generated based on the described information.The components and assemblies of the plant can represent nodes in the knowledge graph and be linked to each other via edges according to the hierarchy of the plant.

[0018] Advantageously, the relevant information is therefore structured in the knowledge graph, particularly at the correct hierarchical level. Furthermore, relationships between assemblies and / or components are advantageously known and can be output in the digital twin of the plant, for example, to a user.

[0019] In a further development of the procedure, the knowledge graph may include a mechanical graph describing the mechanical relationships of the assemblies and / or components, an electrical graph describing the electrical relationships of the assemblies and / or components, and / or a structural graph describing the spatial relationships of the assemblies and / or components. As previously described, the relationships between the assemblies and components of the plant can be determined based on the plant information using artificial intelligence, in particular generative artificial intelligence. Based on this, the knowledge graph described in 25-0645 DE can be generated.

[0020] The embodiment may include a mechanical graph and / or an electrical graph and / or a structural graph. The structural graph, for example, may not be a separate graph, but may be included as additional information within the mechanical graph.

[0021] The mechanical diagram describes, for example, the mechanical connections and relationships between the individual components and / or assemblies of the system. This can include, for instance, which components or assemblies are mechanically connected or interact with each other. The mechanical diagram can be generated, for example, based on bills of materials, quantity take-off files, and similar documents.

[0022] The electrical diagram describes the electrical relationships of the assemblies and / or components. This means that the electrical diagram specifically describes or encompasses how the individual components of the system are electrically connected. The electrical diagram can be generated, for example, based on at least one circuit diagram, or the system information underlying the electrical diagram can be derived from circuit diagrams and similar sources.

[0023] The structural diagram describes the spatial relationships of the assemblies and / or components. This information can be obtained, for example, from a coordinate system of the plant or from three-dimensional models. In other words, the structural diagram can describe the spatial arrangement of the assemblies and / or components.

[0024] In a further embodiment of the method, the algorithm can determine, based on the plant information, in particular based on names and / or shapes and / or connections, an assignment of components and / or assemblies into the hierarchy and / or a mechanical and / or electrical or spatial relationship of the components and / or assemblies. As described, the algorithm receives the plant information, in particular in an unordered and unstructured form, in various formats and from various sources. The algorithm assigns the components and / or assemblies into the hierarchy or hierarchical structure of the plant based on the plant information. This can be done, for example, based on names and / or shapes.

[0025] and / or connections of the individual components and assemblies are described in the plant information.

[0026] For example, the algorithm searches plant information, particularly parts lists, manuals, and the like, for names and determines relationships based on the names used. Similarly, shapes and / or connections, such as three-dimensional forms and / or electrical or mechanical connections, can be found in the plant information, for example, using three-dimensional models, circuit diagrams, and the like. This means that the algorithm can locate assemblies and / or components in different sources of plant information, such as various documents, models, lists, and the like, and thus determine the relationships within the plant, for example, which assemblies and components are mechanically connected and which components are electrically connected, and how these connections interact.For example, the algorithm can locate a component in bills of materials, circuit diagrams, three-dimensional models, and the like, and thus determine the component's relationships to other components or assemblies. Based on this, the previously described information can be generated, in particular hierarchy information and relationship information.

[0027] The process can also include assigning components and / or assemblies in the plant information based on a certain probability. For example, the algorithm can identify identical or similar designations in the plant information. The most probable matches can then be assigned to each other. If, for instance, the designations for components and / or assemblies in the bill of materials or a circuit diagram are identical or match with a sufficient probability, the previously described relationships can be automatically determined by the algorithm. For example, an assignment can be determined based on a similarity analysis, particularly if the designations are not identical but similar, especially if they share the same identifier.In this process, the most probable "matches" of the designations in the individual sources of plant information can be assigned to each other. 25-0645 DE.

[0028] In a further embodiment of the method, the assignment of an assembly and / or a component can be verified or modified based on user input and / or a query to an external data source, particularly the internet. As described above, the assignment of assemblies and / or components to one another can be carried out spatially, structurally, electrically, or based on a data connection, for example, based on a certain probability, or the hierarchical structure of the system can be determined. User input and / or a query to an external data source can be used for this purpose, particularly for verification or correction.If the assignment of a specific component or assembly is ambiguous or insufficiently similar, the assignment can be verified by searching an external data source, such as the internet. Alternatively, the assignment of the assembly and / or components, for example, within a subsystem of the plant, can be determined and verified using manufacturer data in an external database. Similarly, a specific assignment can be presented to the user for review, allowing them to verify or modify the assignment based on their input.

[0029] If conflicts or errors are identified, for example, due to discrepancies in information from different sources of plant data, a detected error, particularly a contradiction, can be displayed in the plant data and / or corrected, especially based on user input and / or a query to an external data source. For example, if errors exist in the provided plant data, such as errors in circuit diagrams, parts lists, and the like, these can be identified and corrected in the plant data, i.e., in the source file or document, or by research in external data sources and / or based on user input. Advantageously, this reduces the workload for a user to the task of checking such contradictory or inconsistent data.ambiguous assignments, so that other assignments, for example of identically found components and / or assemblies, can be carried out automatically.

[0030] The provided plant information can be preprocessed in a further embodiment using at least one preprocessing algorithm, 25-0645 DE

[0031] In particular, data can be extracted and transformed into a uniform format. The preprocessing algorithm can, for example, include optical character recognition (OCR) or pattern recognition to search documents, especially text documents or image data, and extract the data that can form the basis for determining the relationships between components and / or assemblies. The preprocessed data—that is, the data extracted from the plant information by the preprocessing algorithm and transformed into a uniform format—can then be enhanced and fed into the previously described algorithm, especially comprehensive generative artificial intelligence, to determine the relationship and hierarchy information.

[0032] The process can further include the assignment of data records to individual components and / or assemblies based on the generated structure. In other words, various pieces of information can be assigned to individual components and / or assemblies within the hierarchical structure of the system, particularly in the knowledge graph. This improves the structuring of the entire dataset, thereby enhancing the overview and usability for the user. For example, if the user requires information about a specific component and / or assembly, assigning data records to the individual components and / or assemblies allows for quick retrieval and structured access to the data records. The user can, for instance, select the corresponding component and / or assembly in the hierarchical structure of the knowledge graph, and the data records will then be stored there.Specifically, all relevant information for a particular component and / or assembly can be stored at the corresponding hierarchy level. The user can therefore retrieve the required information at the specific location in the knowledge graph, for example, a node associated with the component or assembly.

[0033] Furthermore, execution or output in the form of a knowledge graph allows for faster searching based on the node structure of the knowledge graph. For example, only the relevant node can be checked, or interconnected nodes can be included in the search based on previously determined relationship information. Additionally, [25-0645 DE]

[0034] that the data structure in itself may differ for different components and / or assemblies or hierarchy levels, depending on which components or assemblies are assigned there.

[0035] Furthermore, the process can include the output of a data record for at least one component and / or assembly based on a user query, particularly via a chatbot, and / or a database query. For example, the user can formulate a query, such as to a chatbot, which can be formulated in natural language. The corresponding data record can then be output to the user, for example, information relating to a requested component and / or assembly. Alternatively or additionally, it is also possible to output a database query. The described digital twin can therefore be understood and used as an interface for various programs or applications.

[0036] Specifically, the described method can be used to build, modify, maintain, and / or control at least one component of the plant based on the knowledge graph. As described, the digital twin of the plant can be implemented as an interface for at least one program or application. Based on the knowledge graph, the plant or at least one component can be built. Maintenance and modifications can also be performed, for example, using the plant's hierarchy and / or relationship information. In principle, the individual data stored in the knowledge graph can be used for such tasks, particularly building, modifying, and maintaining the plant. It is also possible to control the plant based on the data stored in the knowledge graph.At least one function of the system, for example the operation of a specific component, can be controlled based on the knowledge graph.

[0037] In addition to the method, the invention relates to a control device for generating a digital twin of a plant, comprising several assemblies with multiple components, wherein the control device is configured to receive plant information describing the plant, in particular from various sources and / or in various formats, wherein the 25-0645 DE

[0038] The control device is designed to generate the digital twin as a knowledge graph using at least one artificial intelligence algorithm, which describes hierarchy information describing the hierarchical structure of the assemblies within the plant and relationship information describing the relationship of the assemblies and / or components within the plant.

[0039] The control device is specifically designed to execute the previously described method in all its details, or the previously described method can be executed with such a control device.

[0040] All advantages, details, designs and / or features described in relation to the method are fully transferable to the control device and vice versa.

[0041] The invention is explained with reference to exemplary embodiments and the figures. The figures are schematic representations and show:

[0042] Fig. 1 shows a schematic representation of a control device according to an exemplary embodiment; and

[0043] Fig. 2 shows a schematic representation of a flowchart of a method for generating a digital twin of a plant according to an exemplary embodiment.

[0044] Fig. 1 schematically shows a control device 1 configured to carry out the method described below with reference to Fig. 2. The control device 1 is configured to execute an artificial intelligence algorithm, in particular generative artificial intelligence. The algorithm generates a knowledge graph 2 to which plant information relevant to a plant can be fed. The plant information can be supplied, for example, from one or more sources 3. Optionally, the control device 1 includes a preprocessing unit 4 that preprocesses the information. 25-0645 DE

[0045] The control device 1 also has a user interface 5 and a program interface 6, through which, for example, a user or at least a program can access the knowledge graph 2. For example, a plant described by the plant information can be digitally represented using the digital twin described by the knowledge graph 2. The plant can be operated, set up, modified, or maintained using the knowledge graph 2, as described below.

[0046] The process for generating the digital twin starts, for example, in Fig. 2 with a block ? in which the plant information is provided to the control device 1. The plant information can include all information available for the plant's assemblies and components. This includes, for example, operating instructions, maintenance manuals, parts lists, three-dimensional models, circuit diagrams, and the like. The plant information is available in source 3 in an unstructured format, with each individual data source or data record potentially being in a different format. For example, the various components and assemblies are supplied by different manufacturers, who provide parts lists, operating instructions, maintenance manuals, three-dimensional models, circuit diagrams, and the like in different forms or formats.

[0047] Optionally, preprocessing can be performed in Block 8, in particular using the preprocessing unit 4 described above. This can include, for example, pattern recognition and / or optical character recognition to identify text, names, or the shapes of components in the various data sources. This makes it possible to preprocess the plant information and use it as the basis for further processing.

[0048] Block 9 allows for the corresponding assignment of information. The algorithm generates a hierarchical structure of the plant based on the provided plant information. This hierarchical structure can comprise several levels, grouping individual components into assemblies, and these in turn into machines within the plant. The hierarchical structure is also determined primarily by the plant information, for example, by the assignment of individual components to assemblies or individual components.

[0049] Assemblies for machines within the plant. Furthermore, relationship information is defined that describes the relationship between the assemblies and / or components within the plant. This relationship information describes, for example, which components or assemblies are mechanically, structurally, electrically, or data-wise connected to each other.

[0050] The assignment can be automated, for example, based on a specific probability. If the algorithm finds identical designations in the plant information, the assignment of components and assemblies can be automated. Alternatively, an assignment can be made via similarity analysis, where the similarity determines, for example, the probability that the assignment is correct. Specifically, the most probable matches can be automatically assigned to each other.

[0051] For example, it is possible to determine which components in a bill of materials correspond to each other in a circuit diagram or a three-dimensional model and how they are related. Accordingly, the information can be assigned to the structure of knowledge graph 2 in Block 9.

[0052] If errors or inconsistencies exist in the individual data records of the plant information, or if automated determination or assignment is not possible, for example, due to a lack of similarity in the designations across different data records, verification or correction can be performed in Block 10. The correction can, for example, be carried out automatically in the plant information, such as a source file. For this purpose, the algorithm can search at least one external data source, such as the internet, for a component and its structure, and determine how it should be integrated into the corresponding plant. Similarly, a possible assignment can be displayed to the user, allowing the user to confirm or correct it. Advantageously, the user is only shown those assignments that cannot be performed automatically. This significantly reduces the user's workload.25-0645 DE.

[0053] Block 11 can provide output, for example, via the interfaces described above, specifically user interface 5 and program interface 6. User interface 5 can, for example, generate output in the form of a chatbot. By connecting control device 1 to other programs or machines of the plant via program interface 6, further functions can be executed, in particular the plant can be set up, modified, maintained, or controlled.

[0054] As described, the control device 1 is designed to execute the method described with reference to Fig. 2, or the method described with reference to the schematic flowchart in Fig. 2 can be fully executed on the control device 1.

[0055] All the advantages, details, and features described in relation to the embodiments shown in Figures 1 and 2 are fully combinable, interchangeable, and transferable. 25-0645 DE

[0056] REFERENCE MARK LIST

[0057] 1 Control device

[0058] 2 Knowledge graph

[0059] 3 Source

[0060] 4 Preprocessing equipment

[0061] 5 User interface

[0062] 6 Program interface

[0063] 7-11 Block

Claims

25-0645 DE REQUIREMENTS 1. Method for generating a digital twin of a plant comprising several assemblies with several components, wherein plant-describing plant information, in particular from different sources (3) and / or in different formats, is provided, characterized in that the digital twin is generated as a knowledge graph (2) by means of at least one artificial intelligence algorithm, which describes hierarchy information describing the hierarchical structure of the assemblies within the plant and relationship information describing the relationship of the assemblies and / or components within the plant.

2. Method according to claim 1, characterized in that the knowledge graph (2) comprises a mechanics graph describing the mechanical relationships of the assemblies and / or components and / or an electrical graph describing the electrical relationships of the assemblies and / or components and / or a structural graph describing the spatial relationships of the assemblies and / or components.

3. Method according to claim 1 or 2, characterized in that the algorithm determines, based on the plant information, in particular based on names and / or shapes and / or connections, an assignment of components and / or assemblies into the hierarchy and / or a mechanical and / or electrical and / or spatial relationship of the components and / or assemblies.

4. Method according to claim 3, characterized in that components and / or assemblies are assigned in the plant information based on a certain probability.

5. Method according to one of the preceding claims, characterized in that an assignment of an assembly and / or a component is verified or changed based on user input and / or a request to an external data source, in particular the Internet. 25-0645 DE 6. Method according to one of the preceding claims, characterized in that a detected error, in particular a contradiction, is output and / or corrected in the plant information, in particular based on a user input and / or a request to an external data source.

7. Method according to one of the preceding claims, characterized in that the plant information is preprocessed by means of at least one preprocessing algorithm, in particular data is transformed into a uniform format.

8. Method according to one of the preceding claims, characterized in that data sets are assigned to the individual components and / or assemblies based on the generated structure.

9. Method according to one of the preceding claims, characterized in that a data set for at least one component and / or assembly is output based on a user query, in particular by means of a chatbot, and / or a database query.

10. Method according to one of the preceding claims, characterized in that at least one assembly of the plant is built and / or modified and / or maintained and / or controlled based on the knowledge graph (2).

11. Control device (1) for generating a digital twin of a plant comprising several assemblies with several components, wherein the control device (1) is configured to receive plant information describing the plant, in particular from various sources (3) and / or in various formats, characterized in that the control device (1) is configured to generate the digital twin as a knowledge graph (2) by means of at least one artificial intelligence algorithm, which describes hierarchy information describing the hierarchical structure of the assemblies within the plant and relationship information describing the relationship of the assemblies and / or components within the plant.