Device for data exchange in a system, and computer-implemented method
The 'need-to-know' data standardization approach with an anchoring mechanism facilitates efficient data exchange between heterogeneous systems by establishing a common domain model, reducing integration complexities and costs, and enabling seamless collaboration across different environments.
Patent Information
- Application Number
- PCT/EP2023/063005
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-15
- Publication Date
- 2024-11-21
AI Technical Summary
Current systems face challenges in integrating data from heterogeneous environments due to varying protocols, standards, and information models, leading to complex and costly data exchange between different systems, especially across the shop floor and cloud contexts, with no established synchronization mechanisms and high integration efforts.
A data standardization approach based on a 'need-to-know' basis, utilizing an 'anchoring mechanism' to establish a common domain model and protocol, enabling interoperability between systems by defining a minimal set of fundamental building blocks and allowing systems to share data without requiring a common 'world model', using the Anchor Core Model and Anchor Domain Models for efficient data exchange.
This approach minimizes redundant engineering efforts, avoids inconsistencies, and enables seamless data exchange across systems, allowing them to collaborate without needing to understand each other's domain details, while maintaining system-specific freedom and flexibility.
Smart Images

Figure EP2023063005_21112024_PF_FP_ABST
Abstract
Description
[0001] Description
[0002] Device for data exchange in a system, and computer-implemented method
[0003] The invention relates to a common, agreed domain model, a data model and a protocol to provide out-of-the-box interoperability between different systems (e.g., even from different manufacturers) in a common context defined on the basis of a knowledge need.
[0004] Today, both shop-floor and top-floor environments are highly heterogeneous, with many different protocols and standards for exposing and sharing data. Data exchange is complicated not only by the multitude of different interfaces, but also by the diversity of information models and the semantic maturity of the data underlying these interfaces: from unstructured raw data without any context to various standards or proprietary data models structured and enriched according to an explicit ontology. Complexity, scope, and usage context vary in many respects.
[0005] Consider the following example: A customer wants to combine data in a cloud integration platform (a software-as-a-service offering in the context of the Internet of Things that stores operational data and makes it accessible via digital applications) and an edge solution (an open, ready-to-use edge computing platform consisting of edge devices, edge apps, edge connectivity, and an application and device management infrastructure), i.e., a context that extends from the shop floor to the cloud. This means that data originating from industrial robots, for example, must be able to be moved across the various layers to make it available for analysis in a cloud solution. Even though the context and the underlying model (in this case, IoT assets) are very similar, several problems arise:
[0006] • The metamodels on which the various data models and standards are based, on which in turn the data of the various robots are based, differ from each other in terms of both structure and features.
[0007] • APIs and other interfaces vary.
[0008] • Both the cloud and edge platforms are "inward-facing," meaning there are no established synchronization mechanisms with the other system.
[0009] It turns out that the two systems cannot simply be plugged together, and the integration effort and costs would be very high.
[0010] There are other examples where a Distributed Control System (DCS) or Supervisory Control and Data Acquisition (SCADA) system needs to be integrated with a Manufacturing Operations Management (MOM) system, or the customer wants to integrate data from multiple shopfloor sources to optimize the process, but no established solution can be provided so far.
[0011] Today, there are many tools available for creating custom solutions that address such integration problems. These solutions are often expensive, time-consuming, and non-reproducible, as they each address unique, highly customer-specific integration problems.
[0012] State of the art
[0013] EP 3084628 describes a method for integrated data processing in a plant constructed and operated by a plurality of software applications with different functionalities, comprising steps for interconnecting these software applications via a communication framework and mechanism, providing a common data model including data shared by at least two of these applications and managing the shared data.
[0014] Approaches to addressing the challenges described can be divided into the following:
[0015] - Leave the data source "as is" and integrate the data using the traditional Extract - Transform - Load procedure.
[0016] - Standardization of the way software systems structure, name, and make data available, usually on a domain basis (one standard per domain type). Within the framework of the standardization approach, a distinction can also be made as to how standardization works:
[0017] - whether the naming, meaning and structure of data elements is prescribed and disclosed down to the smallest detail of a specific domain or
[0018] - whether the enforcement of standardization is kept to a minimum (data standardization on a need-to-know basis, standardization of only the common context).
[0019] In the following, various well-known approaches are described in more detail.
[0020] Option A: Full-depth point-to-point data integration, 70
[0021] Traditionally, attempts have been made to solve the problems of integrating heterogeneous data described above by establishing an additional layer for point-to-point data integration, with all the necessary connectors for data extraction and transformation from the various, individual data sources at any level of depth, or alternatively by means of virtualization. However, an additional layer is not always desirable or useful from a customer perspective, for example, in cases where the customer only wants to exchange data bilaterally between two individual tools in a traditional one-to-one process.
[0022] Furthermore, the number of integration mechanisms grows with each data source added to the point-to-point integration. An integration layer also shifts configuration effort to a solution business or results in additional engineering effort on the customer side.
[0023] For Option A, there are already data integration tools and solutions on the market, for example:
[0024] • Cognite Data Fusion® Platform, https: / / www.cognite.com / en /
[0025] • Element Unify, https: / / www.elementanalytics.com /
[0026] Figure 7 shows Option A, the interoperability of, for example, three systems 71, 72, 73 via a 1:1 integration layer 70. As can be seen from the illustration, for three systems, six connections 712, 721m, etc. are already required. For four systems, 12 translation connections are already required.
[0027] Option B: Comprehensive data standardization for each domain, 80
[0028] Instead of the traditional heavyweight data integration of Option A, with all the costs associated with highly customized data integration, Option B standardizes the way data is structured, described, and made available from its source. Ideally, this leads to interoperability (by design), so that data can be more easily consumed and exchanged via common APIs and serialization formats. The advantage of this option is the complete transparency and predictability of data transfers provided by standardization.
[0029] Possible ways to achieve this are either the standardization of the underlying data models themselves or the data models remain diverse but refer to a common metamodel for data.
[0030] However, the existing standards landscape is fragmented and is usually not sufficient to completely solve a customer problem.
[0031] The standards typically address only a specific area of interest and often do not refer to a common metamodel, making the translation of the data models a difficult task. Furthermore, in the case of new functions, existing systems or standards will not cover the new aspects.
[0032] It also prescribes how and with what level of detail a particular software system should describe, structure, and make its data available. This makes the system development process very inflexible and unconstructively limits the range of design options available to a system architect.
[0033] In addition, each introduction of a new standard with which the various software systems are to be compatible requires separate translation efforts for each of the systems.
[0034] For Option B, the following standards, standardization approaches and modeling languages can be found in the state of the art:
[0035] • AssetAdministration Shell (AAS)
[0036] • OPC-UA Companion Standards
[0037] • Siemens Ontology Library
[0038] • Bosch Aspect Metamodel (BAMM) Microsoft DTDL Network of Things ISO 15926-14
[0039] Some standards, such as BAMM and AAS, focus on defining a common metamodel with a serialization format to exchange descriptions based on these metamodels in a standardized manner. Furthermore, these standards focus on selected domains, such as automation or device descriptions, and therefore cover only a portion of the desired solution space.
[0040] Figure 8 illustrates an example of Option B. Each of the three systems 71, 72, 73 is compatible with one of the standards 81, 82, 83, for example, ISA-95, VDI 2770, or AAS. Each system includes its own local translation solution 84, 85, 86, which includes heuristics 843, 842, 853, 852, 863, 862, ..., which contains instructions for translating its own standard 81 (or 82, 83) into the respective other standards 82, 83 (or 81).
[0041] Here, too, the number of heuristics required increases significantly with the number of underlying systems.
[0042] The object of the invention is to provide a method which solves the described problems of system integration and enables data exchange between different systems in a plant.
[0043] The object is achieved by a device having the features according to patent claim 1.
[0044] The problem is also solved by a method having the features of independent method claim 6.
[0045] The object is further achieved by a computer program product according to the features of patent claim 13. Advantageous embodiments of the invention are described by the subclaims.
[0046] A data standardization approach is proposed that operates on a need-to-know basis with a focus on a common context.
[0047] The proposed approach overcomes the weaknesses of options A and B described above while making use of their respective strengths.
[0048] By adding a so-called "anchoring mechanism" to existing, completely independent systems (e.g., to a robot and to a machine tool in a machine tending scenario), we enable a basic form of interoperability that allows the systems to exchange a predefined set of information (e.g., material IDs that a robot should deliver to the machine tool).
[0049] The purpose of the proposed model is to provide a common, agreed-upon domain model, a data model, and a protocol to enable out-of-the-box interoperability between different systems (e.g., even from different vendors) within a common context defined based on a "knowledge need." This enables more efficient integration between products.
[0050] Such systems can be, for example, different software products, different assets, or any combination of both.
[0051] The subject matter of the application, as well as further embodiments, are explained below by means of the figures.
[0052] Figure 1 is a schematic representation of the model,
[0053] Figure 2 Multi-Layer Anchor Model
[0054] Figure 3 Type Level Engineering
[0055] Figure 4 Instance Level Engineering
[0056] Figure 5 Example scenario
[0057] Figure 6 Comparison with the state of the art
[0058] Figure 7 State of the art, Option A: Integration layer,
[0059] Figure 8 State of the art, Option B: with standardization
[0060] Figure 1 shows a schematic representation of the concept. It again depicts three systems (11, 12, 13), which are compatible with various standards (81, 82, 83), such as ISA95, VDI2770, and AA8. Each of the systems includes an Anchor Advanced Domain Model (AAD) with a specific focus, for example, on the asset hierarchy (111), on product data (121), or on the product's CO2 footprint (131).
[0061] When systems can collaborate using a common model supported by all participating applications, these applications do not need to be manually integrated 1:1 (see Scenario A described above). Instead, they are able to collaborate and exchange data more automatically and seamlessly, minimizing redundant engineering effort and avoiding cross-system inconsistencies. At the same time, it avoids the need for each of the participating systems to be able to understand the other, especially system data in all domain details, but only on a "need-to-know" basis, aligned to a common context.
[0062] The term "anchor", also used below, refers to the concept, 10, that participating systems (i.e., products) do not need to have a common "world model", with the intention of defining every detail in this large model and thus having to consider all requirements of all products in all details of the model.
[0063] Rather, a mandatory set of basic building blocks is specified at the metamodel level, 19 which allows applications to create their own objects from them. However, because these objects are modeled with a minimum set of common entities, attributes, and relationships, applications can view these objects as a "secure anchor" between themselves and other systems. These "anchors" enable collaboration by identifying and using base objects of the anchor model. It is also possible to use information outside the anchor model scope using external references that link semantic models, ontologies, and taxonomies in applications that are able to follow these references.
[0064] The specification provides a common metamodel (the Anchor Core Model) 19 and, based on it, very simple anchor domain models, 18 (assets, process objects, ... ), which can be extended to define advanced domain models .
[0065] Definition Anchor Registry, 16
[0066] - Keeps track of all available assets and where to find them
[0067] - Can translate between different IDs of assets in different domains (system-specific, external, unique,...)
[0068] - Creates a unique namespace, optionally in combination with a central broker (managing topologies and data flows)
[0069] - Optionally, a distributed approach without registers is conceivable.
[0070] Definition Type Repository, 17
[0071] - Tracks all defined anchor types and their history - Allows central sharing and browsing of type definitions without having to redefine types within each system
[0072] - can be pre-filled with known types or known information models
[0073] - Expandable by all
[0074] Definition of Anchor Domain Models, 18
[0075] Each anchor domain 18 has a generic API for interacting with the anchor model, but no way to define methods and events for custom entity types. Its primary purpose is to expose, describe, and manipulate data without losing semantics. It does not allow the definition of operations for interacting with custom objects (e.g., calibrating a robot or starting a machine).
[0076] The anchor model has extension mechanisms that allow linking to external models, instances, and even interfaces.
[0077] Figure 2 shows in more detail how these different artifacts are linked.
[0078] The basis is again the Anchor Core Model, 19 a metamodel with a toolbox for creating all possible models, defining rules and possibilities. This model is kept as small as possible to minimize complexity.
[0079] Building on this are the Anchor Domain Models, 18, also very lean models that cover the needs of the various domains regarding how data is handled, with regard to organization, inventory, exchange, and interaction. Multiple Domain Models 18 can also be supported in parallel. For example, a distinction can be made between an Asset Anchor Model, 25, and one (or more) Process Anchor Models, 26.
[0080] Above this is a layer called the "Advanced Domain Model" (AAD), 22. These models relate to specific products and can also include and combine elements from different domain models, 18. For example, the ISA-88 Anchor Model, 24, is included.
[0081] Last but not least, there is a layer of customer-specific models, 21, which are created and maintained for specific customer requirements, 23.
[0082] Data interfaces are either product-specific 27, or of a general nature 28.
[0083] Products can enrich APIs with product-specific features, such as advanced queries, which are then not supported or used by other products.
[0084] Instead of just creating one model for linking different standards with different models and serialization formats, we stay within the same metamodel to have the same rules for identification, structuring, and relationships across all categories and external references.
[0085] On this basis, data exchange between two individual systems is permitted. This particularly concerns interoperability.
[0086] Collaboration and data exchange across multiple systems in a common model is offered. This allows systems to contribute to this model and / or enrich it with additional information.
[0087] Two different, situation-dependent approaches to the application can be distinguished: During engineering:
[0088] Method of configuration and commissioning, i.e. anchoring a connection between two different data systems.
[0089] During the term:
[0090] New incoming data is automatically transformed between the systems.
[0091] In the following sections you will find further information and implementation examples of the proposed solution.
[0092] Procedure for configuring and commissioning (ie "anchoring") a connection between two different data systems
[0093] To build such a multi-layered model, an engineering process is proposed that enables data to be anchored consistently across distributed systems at both the type and instance levels.
[0094] Two roles are introduced for coordination between the different systems:
[0095] - a first register, the so-called "Anchor Type Repository, 30 and
[0096] - a second register, the "Anchor Registry", 16.
[0097] The roles can be taken over by one of the systems (e.g.
[0098] B. integrated) or through an independent 3rd party system.
[0099] The roles are described based on Figure 3 and Figure 4 .
[0100] The anchor construction takes place on two levels:
[0101] Type-level engineering (as shown in Figure 3) based on the data models of the respective system, 114, 124, with the following steps: 1. Identify the relevant anchor domains (e.g., equipment, capabilities, order, location) based on the data model of System A, 114, 124, and identify the available base types to be used (provided or retrieved). Base models are retrievable from the Anchor Type Repository, 30.
[0102] Adding additional base types is not permitted.
[0103] 2. Selecting additional types from potentially available extended domain models, 115, 125, as well as from already created custom models, 116, 126 (see also Figure 2). Extended and custom models can be retrieved from the Anchor Type Repository, 30.
[0104] 3. If additional domain objects with well-defined semantics are to be exchanged that do not yet exist as types in advanced or user-defined models, the creation of new types can be initiated via the Anchor Type Registry, 117, 127.
[0105] The Anchor Type Repository ensures consistency through a data governance process that depends on the organizational setup, the business scenario, etc., and is therefore not detailed here.
[0106] 4. Based on the defined anchor type models, the API can be derived and provided as an additional, standardized API interface to System A. 118, 128.
[0107] For additional systems B...Z, the engineering process is repeated at the type level. It should be noted that extensions to the extended and customer-specific models triggered by one system may potentially lead to a reimplementation of the other systems.
[0108] Therefore, all extensions must be carefully regulated as mentioned in step 3 above.
[0109] B) Engineering at instance level (shown in Figure 4): 1. Entity search:
[0110] Retrieving existing entities from the local system and matching the entities with the entities in the Anchor Registry, 16, based on IDs and attributes / relationships required for the corresponding type.
[0111] Depending on the anchor registry, different forms of entity matching and deduplication algorithms (i.e., detecting and eliminating redundant data) can be used.
[0112] 2. If the entity already exists in the anchor registry, the available entity definition (all available relationships, attributes, references) is downloaded and compared with the local entity. 412. In case of differences, a predefined merge routine is applied (e.g., overwriting local definitions). Local data structures cannot be easily modified, so this scenario typically leads to local mapping procedures that allow transforming local instances into anchor-compliant instances. 413.
[0113] If the entity does not exist in the anchor registry, a new anchor entity is registered, 414, via a registry API.
[0114] 3. Optionally, the entity description can be extended by additional relations, attributes or references, which leads to additional assignments.
[0115] 4. All mappings developed in steps 2 and 3 must be executed at runtime when data is read or written through the anchor interface. Therefore, the mapping logic must be implemented and distributed as part of the anchor interface.
[0116] Mappings can be represented and executed in several ways. For example, a simple mapping of identifiers can be achieved using a mapping table that assigns an anchor ID to each internal entity ID. However, in most cases, more complex mappings are required, including data structure transformations.
[0117] An exemplary embodiment is shown in Figure 5: The following describes an example situation in which robots from different manufacturers must be able to match tasks from a task management system to the correct milling machines 11, 12, depending on the milling machine requirements: orientation and hardness of the available cutters. Only the type of information that (must) be shared in this specific area is expressed as an anchor (need-to-know):
[0118] - Specific features of a milling task, but not of other types of tasks;
[0119] - Specific characteristics of milling machines that are relevant for milling but not for other types of tasks such as milling machine maintenance or onboarding.
[0120] A first domain model, "Milling" 51, in the "Maintenance" area contains information about the device 511, for example, the number of axes or the orientation and hardness of the milling attachments, 512. A further description 513 is included, which refers to other asset domain models, for example, "Common Device Management" 52. This contains data on the device type 521 (description, manufacturer, serial number, 522), the robot type 521 (description, online status 522), and the machine type. A further description 523 is also possible here. The next information relates to task planning 53, which includes, for example, information 531 about the minimum required hardness or the necessary orientation of the tool, 532.
[0121] The challenge here is to enable automated task matching with any number of robots and milling machines, each from a different manufacturer and therefore equipped with different, manufacturer-specific interfaces, data structures, and protocols. There is currently no industry standard that defines a commonly agreed-upon data model and consistent ID identifiers.
[0122] The steps described above can be used to add an additional interface to such systems, allowing them to exchange task and capability information in a standardized manner without having to replace the existing data modeling approach and interface. Furthermore, other systems C...Z can also integrate using the same anchor mechanism.
[0123] In contrast to traditional data integration approaches, which require a fully integrated data model, the focus of the approach described above is on the shared context for data exchange between two or more collaborating systems. The modeling focus is limited to a few essential elements, i.e., only those relevant to the shared context (need-to-know). Within this, the anchor model also specifies an agreed-upon vocabulary for describing these elements. The described approach of the anchor model aims to balance the need for system-specific freedom to use "deep," domain-specific data models without having to adhere to a common, fully integrated data model in every detail, and the need to avoid 1:1 integration efforts between individual systems.Figure 6 shows the described concept 60 in comparison with the concepts of Option A, 70, and Option B, 80, explained under the state of the art. The basis is the existing data from systems A...D, 61, 62, 63, 64.
[0124] Unlike other modeling or standardization approaches, the anchor model offers two combined options. First, organizations can model any domain and any semantics using the anchor model. Second, existing and external representations (based on, for example, OPC-UA, ECLASS, AAS, etc.) can be linked without actually having to move them into the new model itself.
Claims
Patent claims 1. Device for data exchange in a plant, using production domains, which is set up for the production or processing of at least one product, wherein the plant consists of plant components (11, 12) which are to communicate with each other, comprising - a metamodel (19) with basic rules for data exchange that are valid for all production domains, - at least one domain model (18) with rules for a domain category, each of which has its own rules regarding the organization and exchange of data within the domain, - at least one extended domain model (22) with rules for exchanging data is required for the product manufactured in the plant, and - at least one manufacturer-specific model (21) for specifying the data exchange based on the extended domain model (22) of a system component (11, 12).
2. Device according to claim 1, characterized in that the metamodel (19) contains the following definitions: -entities, -attributes, -relationships between the entities and -References of each of the previously mentioned elements to external domain model.
3. Device according to one of the preceding claims, characterized in that the domain model (18) comprises data models for at least one of the Hierarchy of plant components, Product, Energy balance, production process, Events in the facility Capabilities of the plant component - Location.
4. Device according to one of the preceding claims, characterized in that the extended domain model (22) uses standardized data structures or modeling languages, in particular: • AssetAdministration Shell (AAS) • OPC-UA Companion Standards • Siemens Ontology Library • Bosch Aspect Metamodel (BAMM) • Microsoft DTDL • Web of Things • ISO 15926-14 • VDI 2770 • ISA-95.
5. Device according to one of the preceding claims, characterized in that the metamodel (19) is connected to a register (16) for storing and managing the information on basic building blocks of all plant components at metamodel level, as well as associated identifiers in the respective associated domain, and a type repository (17) for storing defined types and their history, for the central release and searching of type definitions without the need to redefine types within each system.
6. Device according to one of the preceding claims, characterized in that it comprises a device for graphically displaying the exchanged information, for asset managers or for checking the conformity of the information.
7. Computer-implemented method for data exchange between domains in a plant using production software, which is used for the production or processing of at least one product, wherein the plant consists of plant components, wherein a plant component comprises software components which are to communicate with each other, wherein - at least one manufacturer-specific model (21) is used to specify the data exchange, which is based on an extended domain model (22) which requires rules for exchanging data for the product manufactured in the plant and which in turn - at least one domain model (18) is used, with rules for a domain category, each of which has its own rules regarding the organization and exchange of data in the domain, which is based on - a metamodel (19) with basic rules for data exchange that are valid for all production domains.
8. Computer-implemented method according to claim 6, characterized in that the metamodel (19) contains the following definitions: -attributes, -Relationships between the attributes and -References from the attributes to the domain model.
9. Computer-implemented method according to one of claims 6 or 7, characterized in that the domain model (18) comprises data models for Hierarchy of plant components, Product, Energy balance, production process, Events in the facility Capabilities of the plant component - Location.
10. Computer-implemented method according to one of claims 6 to 8, characterized in that the extended domain model (22) uses standardized data structures or modeling languages, in particular: • AssetAdministration Shell (AAS) • OPC-UA Companion Standards • Siemens Ontology Library • Bosch Aspect Metamodel (BAMM) • Microsoft DTDL • Web of Things • ISO 15926-14 • VDI 2770 • ISA-95.
11. Computer-implemented method according to one of claims 6 to 9, characterized in that a graphical representation of the exchanged information is included, for asset managers or for checking the conformity of the information.
12. Computer-implemented method according to one of claims 6 to 9, characterized in that the extension of the system is carried out in systems engineering at type level with the following steps: Identify the relevant domains from the metamodel Adding associated manufacturer-specific domains Determine missing domains and create new domain objects in the register, Providing an interface . 13 . Computer-implemented method according to one of the claims 6 to 9, characterized in that the extension of the system is carried out in systems engineering at the instance level with the following steps: Identify existing entities of the extension and Compare with the contents of the register and if the result is positive, use the definition of the entity found in the register or if the result is negative, register the new entity in the register, Extending the entity description with additional necessary information. 14 . Computer program product, suitable and arranged to carry out one of the methods according to one of claims 6 to 12 .
Citation Information
Patent Citations
Method for an integrated data handling for the engineering and operation of a plant
EP3084628A1
