vis IT [ Industrial IoT – Digital Twin ] B S O I r e f o h n u a r F www.iosb.fraunhofer.de ISSN 1616-8240

Editorial Prof. Dr.-Ing. habil. Jürgen Beyerer Dr.-Ing. Olaf Sauer Dear friends and partners of the IOSB, Industry 4.0 and its application on the shopfloor are underway. New buzzwords, such as Industrial Internet of Things (IIOT), Digital Twin and Artificial Intelligence (AI), are becoming common parlance. In this issue of visIT – Fraunhofer IOSB’s inhouse magazine – our authors seek to shed light on some of these new trends and describe how Industry 4.0, IIoT and digital twins fit into the bigger picture. Our guest authors, Jivka Ovtcharova and Rainer Drath, are both experts in digital engineering. They discuss their own R&D-results, important aspects of dealing with digital twins, their weak points and the requirements that they have to fulfill. As they explain, digital twins have to cover issues across the lifetime of the related assets. In Industry 4.0 and its related work- ing groups, the digital twin is linked to the asset administration shell, which is regarded as the collection of meta-models and the communication layer of a physical asset. The authors from our industrial automation site in Lemgo explain how digital twins may be used for alarm management in complex and distributed automation systems. Clearly, digital twins require secure communication, not only for their configuration but especially for their run-time data. Jens Otto and Felix Specht describe one industrial use case which illustrates how companies apply OPC UA and thus fulfill the requirements of IEC 62443. Julius Pfrommer takes a closer look at OPC UA publish/subscribe features and their com- bination with real-time capabilities based on Time Sensitive Networking (TSN). Ljiljana Stojanovic focusses on another aspect of digital twins: the meaning of data and its processing close to the asset, e.g. on an edge device. This ‘semantic edge’ is used to com- municate with the industrial asset, such as a machine, perform local processing (like edge analytics) and transfer pre-processed data to the cloud. Christian Kühnert describes the application of cloud services for machine learning and data analytics. The IOSB has developed this platform because most manufacturing companies do not have the capabilities to analyze large amounts of sensor data: manufacturing companies may use it on their own premises or hosted on IOSB-servers. Thomas Usländer’s article covers the important topic of engineering digital twins. During every phase of an asset’s life cycle, the digital representation must be synchronized as closely as possible with the status of the real-world object. Furthermore, digital twins must support the evaluation of past situations and simulation of future situations by means of prognostic models which are always connected to the physical asset. Thus, digital twins need the deployment and run-time environment to be aware of the status of its asset at all times. Smart factories will be connected worldwide and consist of smart assets and their digital twins. These describe, e.g. manufacturing capabilities, skills and capacity levels. In their article, Byung Hun Song from our Korean partner KETI and Kym Watson explain the concept of a smart factory web, which is simultaneously our testbed in the Industrial Internet Consortium. We hope this edition of visIT gives you a comprehensive overview of the many facets of smart assets and their digital representations, their applications and benefits. Karlsruhe, April 2018 Prof. Dr.-Ing. habil. Jürgen Beyerer Dr.-Ing. Olaf Sauer 3 vis IT Industrial IoT

Fig. 2: Status quo of industrial implementation. combined economic potential of more than € 78 bn by 2025. However, this potential can only be achieved if digital twins are implemented in a comprehensive and self-optimizing manner which enables them to adapt to future changes. Current studies show that mature and widespread implementation has not yet taken place (Fig. 2). There are three main weak points to be discussed: - model semantics of digital twins is mostly geometry driven - analytics are aligned but not embedded into the model - simulation and user-interaction are offline. To overcome current limitations, the following three conditions must be fulfilled (Fig. 3): Fig. 3: Preconditions for mature and widespread implementation. - model semantics must be usage driven and adaptive - analytics should be embedded and work in runtime and automatic operation, and for configu- the practical advantages of the proposed - implementation should go along with ration via intuitive man-machine interfaces solution by increasing productivity more experiments and experiences. (e.g. web surface, haptic interaction devices). than 20 percent (Fig. 1). This makes it easier to take decisions based Implementing digital twins demands that on up-to-date, transparent information. Systems which make it easier to gain a we put real problems “into the sandpit“ Thus, by merging real and virtual environ- unique, deep knowledge of assets and their of business units, think, try out, create ments, intelligent commissioning of pro- behaviors throughout the life cycle will “all-in-one“, apply emerging digital tech- duction can be used to generate forecasts pave the road to achieving new levels of nologies playfully and quickly, test new based on real-time data from the shop- optimization and business transformation. solutions in runtime to gain experience fast floor. How does it work? A ready-to-use For example, we want the physical build and transform knowledge into actions and solution at the Industry 4.0 Collaboration to return data to its digital twin through skills. The time to act is now! We invite you Lab of the Institute for Information sensors so the digital twin contains all the to join us in establishing “German Digital Management in Engineering (IMI) at the behavioral information we would have if Twin Engineering” as a trademark. Karlsruhe Institute of Technology (KIT), we inspected the physical build itself. Germany, offers a good example. The use case of a digital twin of a milling machine According to the German Association for used for process optimization and net- Information Technology, Telecommuni- working in virtual reality, while taking cations and New Media (BITKOM), digital account of resource flows, demonstrates twins in manufacturing industry will have a 5 vis IT Industrial IoT

AAS will become synonymous with the digital twin in its future, fully enhanced version. As soon as the meaning of the digital twin finally morphs into the third interpretation, the differences between it and the AAS will disappear. And what is an Industry 4.0 component? It is the physical asset together with its Asset Administration Shell/digital twin. It is an Industry 4.0 enabled device, that can register itself in the I4.0 network and be identified, explored and processed via Industry 4.0 interfaces. Finally, the evolution of the term “digital twin” offers a powerful illustration of the upcoming revolution. The future digital Fig. 2: Three-level concept of a cyber-physical system [4]. The evolution is ongoing. The term “digital with their individual digital twin whenever twin, in combination with cloud technology, twin” is slowly changing its meaning yet necessary at a later date. apps and algorithms, has the potential again. In a digital future, we will need to revolutionize every aspect of industry digital twins as comprehensive physical and Hence, the digital twin will become a power- because it touches every aspect. The functional models for every physical asset ful electronic data object with interfaces: possibilities are endless, the digital twin is [3], e.g. a component, product or system. it will hold or reference all useful data (see only the beginning. Step-by-step, the digital twin will cover all Fig. 1), some data will be semantically stan- the useful information which is relevant dardized (e.g. properties, geometry, topo- across the lifetime of the related asset, from logy), other data will be of a proprietary the idea to the engineering, logistics, ope- nature (e.g. ABB function blocks). Internally, ration, maintenance, reuse and destruction. the digital twin will communicate with its A future digital twin may contain a simula- physical asset, e.g. via proprietary interfaces. tion model, but also a 3D model, hundreds However, externally, it will communicate via of properties, historical data, handbooks, well-defined Industry 4.0 interfaces. installation guidelines, proprietary function blocks, interlockings, state models, alarm The future digital twin will incorporate both and event definitions etc. Friends of the first data and interfaces and be similar to a soft- or second interpretation may be surprised to ware driver – but far more. It will be a multi- hear that it might even have no simulation facetted digital counterpart of the real asset, model at all, e.g. for static assets. It will be embedded in the Industry 4.0 ecosystem, stored in a future Industry 4.0 infrastructure, access point for a new generation of apps be searchable, explorable, associated with and algorithms, mediator between future and sometimes connected to its real coun- Industry 4.0 services and the real world. terpart. It will not be hidden in proprietary Fig. 2 shows the digital twin in the middle simulation tools. Sales tools, simulation of the tree layer concept of a cyber-physical tools, engineering tools, certification tools, system [4] in an Industry 4.0 environment. maintenance tools etc. may connect to digital twins for sales, engineering, certifi- And what is the Asset Administration Shell cation, maintenance, simulation or optimi- (AAS) [5]? According to [6], it contains the zation purposes, sometimes long before the information and I4.0 interfaces for an asset. related real assets are ordered. It will be Sounds similar to the digital twin, doesn’t possible to associate delivered real devices it? This is exactly what [6] proposes: the Reference: [1] Kagermann, Henning; W. D. Lukas and W. Wahlster: Industrie 4.0: Mit dem Internet der Dinge auf dem Weg zur 4. industriellen Revolu- tion. VDI Nachrichten (2011). [2] M. Shafto; M. Conroy; R. Doyle; E. Glaessgen; C. Kemp; J. LeMoigne and L. Wang: Draft model- ing, simulation, information technology & pro- cessing roadmap. Technology Area, vol. 11, 2010. [3] S. Boschert and R. Rosen: Digital twin the simulation aspect. In Mechatronic Futures. Springer, 2016, pp. 59–74. [4] R. Drath and A. Horch: Industrie 4.0: Hit or hype?. IEEE industrial electronics magazine, vol. 8, no. 2, pp. 56–58, 2014. [5] Struktur der Verwaltungsschale: Fortentwick- lung des Referenzmodells für die Industrie 4.0- Komponente, April 2016. [Online]. Available: [4] https://www.plattform-i40.de/I40/Redaktion/DE/ Downloads/Publikation/struktur-der-verwaltungs- schale.html. [6] C. Wagner; C. J Grothoff; U. Epple; R. Drath; S. Malakuti; Grüner S.; Hoffmeister M.; Zimmermann P.: The role of the Industry 4.0 Asset Administration Shell and the Digital Twin during the life cycle of a plant. In: Proceedings of the 2017 IEEE Conference on Emerging Technologies Factory Automation (ETFA 2017), Limassol, Cyprus, 2017. 7 vis IT Industrial IoT

Figure 1: Gartner Hype Cycle for Emerging Technologies 2017. platforms and tools, from engineering to from operations, pre-process the data after sales services. The following use cases according to VDMA 24582 in CODESYS V3 already show the potential benefits, e.g. and transfer this information to a cloud. for optimizing products and processes almost during runtime: 3. Every HOMAG machine for manufac- turing wooden workpieces has its own 1. The automation supplier WAGO has asset administration shell including a developed an approach called DIMA proprietary XML-description and an (decentralized intelligence for modular OPC UA-communication. Homag offers equipment) which allows modules from machine-related services, e.g. the diag- different vendors to be integrated into a nosis system woodScout. This includes final production system. At its core is the integrated machine documentation via a Module Type Package (MTP), a kind of connector from the Tapio cloud, which is self-description of the modules. The MTP owned by Homag and based on Microsoft can be accessed through an interface Azure. Homag’s customer Nobilia, a and contains communication parameters, producer of kitchen furniture, uses functional production services that the Homag digital twins to demonstrate a module offers to the entire production virtual, customer-specific production system and information for the system for lot size 1. production monitoring system. These initial examples clearly show that 2. Components from FESTO are described the application of digital twins will be as AutomationML models and include very specific and always unique to the use geometry, kinematics and software. They case. However, it must also be possible to also refer e.g. to EPLAN schematic servi- integrate the different part models of the ces, which are FESTO-built macro libraries digital twins easily based on their unique for EPLAN Electric P8, V2. The components description. also store data from the application and Reference: [1] Walker, M.J.: Hype Cycle for Emerging Tech- nologies, Gartner, 2017. Report ID G00314560. [2] Drath, R.: Die Rolle der Industrie 4.0 „Ver- waltungsschale“ und des „Digitalen Zwillings“ im Lebenszyklus einer Anlage. VDI-Bericht 2293, Automation 2017, Düsseldorf, VDI-Verlag 2017, S. 600-615. [3] Sauer, O.: Lösungsbausteine für herstellerun- abhängige, standardisierte Schnittstellen in der Produktion. In: Schulz, Th. (Hrsg.): Industrie 4.0: Potenziale erkennen und umsetzen. Vogel Business Media, 2017, S. 87-106. [4] Pfrommer, J.; Schleipen, M.; Beyerer, J.: Fähigkeiten adaptiver Produktionsanlagen. atp edition 55, Nr. 11, 2013, S. 42-49. [5] Schleipen et.al.: SecurePLUGandWORK. Abschlussbericht. Karlsruhe: Fraunhofer IOSB, 2017, 105 S. http://publica.fraunhofer.de/ dokumente/N-439001.html. 9 vis IT Industrial IoT

Fig. 2: The concept of a digital twin for industrial alarm management: a simulation model generates data used for case base construction, machine learning and analysis to support the operator of the production system. and time-consuming searches for the root ALARM MANAGEMENT USING A DIGITAL TWIN In IMPROVE, the basis for the implemen- the alarm system as well as supporting the tation of a digital twin for industrial alarm operator during plant operation. management is a simulation environment that uses the PhysX engine for discrete This project has received funding from the event simulation. The PhysX engine controls European Union's Horizon 2020 research the behavior of basic physical elements, and innovation programme under grant for example tracks and conveyor belts or agreement No. 678867. plant. Machine learning approaches can find similar, previously seen and annotated cases to suggest possible solutions to the operator [5]. Furthermore, a case base can also be used to learn a variety of further models – such as causality models based on Bayesian networks or Markov chains. Such models shed light on dependencies between alarms, which can be used to reduce the number of alarms displayed. The combination of the simulation model with the embedded behavior and the model-learning and analysis approaches constitutes the digital twin (illustrated in Fig. 2). In the context of industrial alarm management, the digital twin is used to construct a database of alarm floods. This is useful for analyzing the overall behavior of Reference: [1] Boschert S.; Rosen R.: (2016) Digital Twin— The Simulation Aspect. In: Hehenberger P.; Bradley D.: (eds) Mechatronic Futures. Springer, Cham [2] IMPROVE project website: http://improve-vfof.eu/ [3] Health and Safety Executive (HSE). The Explo- sion and Fires at the Texaco Refinery, Milford Haven, 24 July 1994 (Incident Report). HSE Books, 1997. [4] J. Wang; F. Yang; T. Chen and S. L. Shah: “An overview of industrial alarm systems: Main causes for alarm overloading, research status, and open problems,“ IEEE Transactions on Automation Science and Engineering, vol. 13, pp. 1045–1061, April 2016. [5] M. Fullen; P. Schüller and O. Niggemann: Semi-supervised case-based reasoning approach to alarm flood analysis. In: Machine Learning for Cyber-Physical Systems and Industry 4.0 (ML4CPS). Lemgo, 2017. 11 vis IT Industrial IoT cause of the problem. This, in turn, may result in dangerous situations, significant downtime and even irreversible damage, such as the infamous explosion at a Texaco refinery which was found to be caused by a flood of alarms [3]. The goal of an intelligent alarm manage- ment system is to avoid alarm flooding and support the operator if it nevertheless occurs. The main reason for alarm flooding is a flawed alarm system design. As revamp- ing the alarm design of a running system is impossible and stopping production impractical, other solutions are required. Traditionally, alarm management has em- ployed simple methods, such as basic signal and alarm filtering to remove alarms before they are displayed to the operator, or giving operators the option of shelving alarms they consider irrelevant or redundant. For anything more advanced than traditional discrete loads representing manufactured goods. The design of the simulation model is enriched with the custom behavior of production units and sensors, definitions for raising alarms based on sensor values as well as a flood detection system. Discrete event simulation allows the user to greatly speed up the flow of time and observe long-term behavior of the plant model – and record the data. Inducing failures in specific modules allows us to collect a case base of alarm flood samples with semantic annotations. Fig. 1 shows an example of methods, a deep understanding of the such a model, simulating a scenario where system – expert knowledge – is needed. This is either very difficult or too time consuming to obtain. Now, the rise of machine learning a fault induced in one module triggers an alarm flood. A simulated case base is a valuable tool which can be analyzed and and data-driven computational intelligence the results applied to the real plant. The raw allows us to consider more complex and intelligent approaches to alarm flooding – by utilizing a digital twin of the system. case base is itself a model of faulty behavior in the plant and can be analyzed using data mining approaches in order to gain insights Intelligent approaches allow us to either into recurring problems in the plant. More- reduce the number of alarms or assist the operator in identifying the root cause. over, it can be used to directly support the operator during an alarm flood in a running

“The technology transfer from Fraunhofer IOSB-INA in the field of automatic configuration and intelligent networking systems, raised our technology level to the state-of-the-art within a very short time period. In the short term, we will launch appropriate products for new applications. Exploiting innovations in this simple way will sustainably strengthen and expand our region OWL”, concludes Peter Boldt, Head of Development at BOGE. This project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No. 678867. Secure Industry 4.0 communication is implemented according to the IEC 62443 standard by applying security concepts based on zones and communication conduits. For more information go to: www.cybersecurity-owl.de developed an information model for BOGE zones of the compressed air system are, in compressors based on the OPC UA device specification. The information model was deployed to an OPC UA server running on a single board computer to allow easy retrofitting of existing compressors. The OPC UA discovery mechanism allows the turn, part of a higher-level zone. A security gateway permits secure communication between the compressed air network and the other networks of SmartFactoryOWL. Communication between the zones takes place exclusively over OPC UA. The solution compressed air management system to is secured by encryption, authentication and detect connected compressors and auto- authorized user groups according to the matically integrate them into a consistent principle of least privilege. compressed air system. Integration of the Plug-and-Work mechanism The integration of OPC UA into the com- into the research factory SmartFactoryOWL pressed air system of SmartFactoryOWL was successfully applied the concept of secure performed in accordance with the IEC 62443 Industry 4.0 communication to existing standard for implementing secure Industry compressed air systems. This solution con- 4.0 communication. The IEC standard defines the aspects of secure Industry 4.0 communication and describes the require- cept, developed by Fraunhofer IOSB-INA, has helped BOGE automatically and securely to connect compressors to the compressed ments for IT security in industrial automa- air management system. It reduces the time, tion and control systems, in particular the concepts of isolated zones and secure communication conduits. Accordingly, the compressed air management system and the compressor are each defined as a separate isolated zone. The individual effort and costs involved in commissioning and configuration. Reference: [1] Otto, J.; Böttcher, B.; Niggemann, O.: Plug-and-Produce: Semantic Module Profile; In: Dagstuhl-Workshop MBEES: Modellbasierte Ent- wicklung eingebetteter Systeme IV; April 2013 [2] Dürkop, L.; Trsek, H.; Otto, J.; Jasperneite, J.: A field level architecture for reconfigurable real- time automation systems. In: 10th IEEE Workshop on Factory Communication Systems, Toulouse, Mai 2014 [3] Otto, J.; Schriegel, S.; Niggemann, O.: Eine Taxonomie für Plug & Produce. In: Automation 2015, Baden-Baden, Jun. 2015 [4] Otto, J.; Niggemann, O.: Automatic Parame- terization of Automation Software for Plug-and- Produce. In: AAAI-15 Workshop on Algorithm Configuration (AlgoConf), Austin, Texas, Jan. 2015 [5] Otto, J.; Vogel-Heuser, B.; Niggemann, O.: Automatic Parameter Estimation for Reusable Software Components of Modular and Recon- figurable Cyber-Physical Production Systems in the Domain of Discrete Manufacturing. In: IEEE Transactions on Industrial Informatics Volume 14, Jan. 2018 [6] IEC 62443-1-1 Industrial communication net- works – Network and system security –Termino- logy, concepts and models; Jul. 2009 [7] IEC 62443-3-2 Security Risk Assessment, System Partitioning and Security Levels; IEC 62443 Security for industrial automation and control systems; May 2017. 13 vis IT Industrial IoT

selected elements of an OPC UA informa- Kalycito Infotech, a systems integrator for implementation realizes this using a special tion model. In that regard, OPC UA publish/ embedded applications, Fraunhofer IOSB representation for the OPC UA information subscribe returns to the definition of tele- has launched a project to develop OPC UA model: every node in the information model gram formats. But in contrast, the DataSet publish/subscribe. This effort is based on is immutable and cannot be modified. It can be flexibly configured at runtime and the open62541 open source implementation is only possible to replace the entire node this definition can be looked up in the server of the OPC UA standard IEC 62541 (https:// with a modified copy. This replacement uses to understand the semantic meaning of open62541.org). open62541 is developed atomic operations, whereby parallel access values according to their origin in an OPC in the C language and can be used for to the information model is possible without UA information model. resource-constrained embedded applica- the use of semaphores. A time-controlled tions. Five companies from the automation hardware interrupt can then trigger the Within the IEEE 802.1 standards series, the industry, all members of OSADL, are jointly generation of publish messages and the enhancement of Ethernet with real-time funding the project. communication under the name of Time information model cannot be in an inconsis- tent state at the time of the interrupt. Real- Sensitive Networking (TSN) has been pushed The first outcome of the project was a time operations for TSN and the flexible forward during recent years. Some parts of demonstrator for Embedded World 2018. best-effort operations can therefore co-ex- the future TSN standard have already been The integration with current TSN hardware ist in one device. This is an important basis adopted, such as clock synchronization of depends on vendor-specific interfaces and for the main idea of Industry 4.0: to flexibly the participants in IEEE 802.1AS and the configuration tools. Nevertheless, large connect automation systems, to perform reservation of transfer capacity via time parts of the OPC UA publish/subscribe additional configuration at runtime and to slots in IEEE 802.1Qbv. This realizes real-time implementation is reusable and has been integrate runtime control with higher-order communication in parallel to normal Ether- folded back into the open62541 project. functionality and services. net operation. One disadvantage of OPC UA has been the lack of real-time guarantees The key feature of the publish/subscribe due to the use of TCP/IP as a “best effort” implementation by Fraunhofer IOSB is the transport layer. Integration with TSN makes connection between the real-time capable OPC UA publish/subscribe an ideal compan- publisher and the non-real-time normal ion to the client/server interaction with OPC UA server. Any activity in the normal additional end-to-end real-time guarantees. server must not delay the real-time publisher Together with the Open Source in Auto- although both interact with the same mation Development Lab (OSADL) and information model. Fraunhofer IOSB’s Demonstrator Reference: [1] OpenMOS Project Deliverable: Assessment of Current Service Bus Technologies. https://www.openmos.eu/documents/11/D3.2_-_ Assessment_of_Current_Service_Bus_Technolo- gies.pdf 15 vis IT Industrial IoT

To help manufacturing companies take full sources at the bottom to the communica- Fig. 2 shows the real-time data analytics advantage of the IIoT, we have developed tion services that support bidirectional service which correlates sensor data and/or methods and tools to intelligently handle communications between the edge services derived data into more meaningful informa- the IoT sensor data and process data at the and cloud services. The vertical layer services tion for making better and faster decisions. edge of the network and deliver faster in- are realized by the semantic models, which It searches for patterns in a continuous sights [3]. The key innovation is in IoTization ensure a common, shared understanding stream of incoming events in order to through dynamic, multi-modal, smart data is achieved across the existing OT and IT detect situations with minimal delay (i.e. gathering, integration and processing based systems, edge services and various human look for a combination of certain types on semantic technologies. roles. of events to create a higher-level business event). The service consists of: As shown on Fig. 1, we do not focus purely The semantic edge is structured in the on asset data; we take into account data following sub-layers: - an app for managing sensors and from the environment in which the asset patterns, creating patterns using the drag operates and the associated processes - the plug-in/out management layer and drop visual interface by combing and resources that interact with the asset. enables on demand plug-in/out based We consider structured data (e.g. sensor on out-of-the-box connectivity; data), semi-structured data (e.g. inspection reports) and unstructured data (e.g. - the knowledge extraction layer creates corrosion images). The semantic edge is machine-understandable representation events and pattern operators (e.g. filtering, aggregation, time-window, etc.), deploying patterns (e.g. on Raspberry Pi) and displaying notifications; used to communicate with the industrial of the raw sensor data; - a set of components based on the WSO2 asset, perform local processing (e.g. edge analytics) and transfer pre-processed data - the smart integration layer ensures data to the cloud. It can be deployed either on fusion in real-time near to the data a machine or on a gateway, depending on sources by dealing with both syntactic the complexity and resource-consumption heterogeneity and semantic of the services to be used. heterogeneity; Siddhi engine for embedded systems for discovering patterns in real-time, sending notifications or even performing actions (e.g. switch on/off a sensor). The proposed approach is driven by real industrial use cases and includes a proof- Looking at the figure from bottom to top, the components are organized in layers - the intelligent service layer comprises of-concept demonstration for validation the many services for building advanced with real-time data. (with push connectors) starting from data applications suitable for the edge. Fig. 2: Real-time data analytics by “moving“ analytics to the edge. Reference: [1] https://ww2.frost.com/news/press-releases/ innovations-enabled-industrial-iot-accelerate- shift-manufacturing-40/ [2] https://inductiveautomation.com/what-is-iiot [3] IoT edge analytics is transforming manufacturing, https://www.ibm.com/blogs/internet-of-things/ smart-manufacturing-edge-analytics/ 17 vis IT Industrial IoT

Fig. 2: Measurement data are uploaded to the cloud and processes using ML algorithms. Results are written back to plant or sent to a web client for visualization. Fig. 2 outlines the approach. Facilities SUCCESSFUL TESTS FOR MONITORING can send a request as streaming data or DRINKING WATER FACILITIES AND cyclically upload their measurements. For QUALITY ASSURANCE IN POLYSTYRENE streaming data this can be done through PANELS standard protocols, such as OPC-UA [2] or REST, but it is also possible to develop a The developed platform has been success- custom interface since internally the plat- fully tested at the IOSB Test Lab (Fig. 1) in form follows a plug-in based architecture. monitoring drinking water facilities [3] and For the response, there are two possible ways for the operator to obtain the results. First, they can be written back to the local quality assurance in polystyrene panel pro- duction. In the former application, the data are uploaded hourly as .csv files and the results provided within the web-interface. management system. This demands that In the latter, the data are uploaded from the facility provides, e.g. OPC-UA nodes for the complete production batch and the the machine learner. Second, they can be panel quality assurance results provided as visualized using a password protected web a report application. Currently, the web application includes a dashboard, the visualization of a process map with state trajectory and access to historic results from the machine learning algorithms. However, the web client can be adapted to the specific needs of the end user. Reference: [1] Kühnert C.; Montalvo I.: A Generic Data Fusion and Analysis Platform for Cyber-Physical Systems, Machine Learning for Cyber-Physical Systems, 2016 [2] Schleipen, M; Kühnert C.; Okon M.; Henßen R.; Bischoff, T.: MoMo – Mobile Monitoring und smarte Datenanalyse basierend auf offenen Standards, VDI Automation 2015 [3] Kühnert, C; Baruthio M.; Bernard T.; Steinmetz C.; Weber J.: Cloud-based event detection platform for water distribution net- works using machine-learning algorithms; CCWI – Computing and Control for Water Industry, 2015 19 vis IT Industrial IoT

of Things (IIoT) for ubiquitous data trans- still be stored for documentation purposes. and design activities are supported by a mission (machine-to-machine – M2M) as well as cost-efficient and scalable data storage make this type of synchronization possible. Despite the enormous progress made in Platform Engineering Information System technology, today’s digital twins are only capable of encompassing partial aspects of the real object. In Industry 4.0, these (PEIS).PEIS helps to mediate between the possibly conflicting demands of users, product managers, software engineers meta-data sets are logically contained in the and technology experts. A methodology CONCEPT OF DIGITAL TWINS so-called asset administration shell (AAS). of this type is indispensable for engineering However, the term digital twin is misleading. that are determined by application domains, and runtime environment which is service- The AAS distinguishes between sub-models digital twins as they need a deployment oriented and based upon open standard technologies. Although biological twins exhibit many genetic matches and start from a common environment, they evolve and socialize industrial sectors and their standards, and views that are defined by the functional and informational shell of those properties that largely autonomously and independently are required by the intended use cases. of each other, at least with increasing age. Conversely, the “socialization” and lifetime ENGINEERING METHODOLOGY OF IOSB of a digital twin in a digital factory is basi- cally independent of that of its real-world Hence, when considering the engineering counterpart. In design and engineering of digital twins, it is important to know departments, digital representations are which properties are required and how to created in virtual environments which are map them to the sub-models. Fraunhofer becoming ever closer to reality. A robot arm IOSB provides a methodology known as may be simulated in its cooperation with SERVUS (service-oriented design of infor- other robot arms or humans and optimized in its behavior before the physical robot arm is even produced and installed. In the mation systems based upon use case specifications) which systematically supports the information system analyst and designer virtual world, time may be wound back and in this task. Validated in industrial project forth as required. This enables analysts to situations, it allows an analyst to specify evaluate past situations and simulate future and document use cases as semi-structured situations by means of prognostic models. tables, break them down into requirements Long after the life of the physical object and map them in an agile manner to has come to an end, e.g. due to wear or capabilities and technologies of existing dismantling, the virtual representation may and emerging IIoT platforms. These analysis SERVUS Meta-model Concept of a digital twin (simplified) 21 vis IT Industrial IoT

The Korean and German model factories will connect to the Microsoft Azure cloud platform to show factory data in real-time. The shop floor data to be transferred to the cloud is aggregated on an OPC UA aggregation server with the help of an information model defined in AutomationML Towards a Marketplace for Manufacturing. and CEP (Complex Event Processing) to extract actionable events from patterns in the data. The AutomationML model is also used to configure the data visualization on the Azure cloud. KETI’s Ansan model factory is also connected to the Siemens’ Mindsphere platform to demonstrate product independence. STANDARDS The standards AutomationML (IEC 62714) and OPC UA (IEC 62541), complemented by the companion specification “OPC Unified Architecture for AutomationML“, play a key role. The combination of these standards reduces the manual engineering effort required for production adaptation and exchange of information in factories. Integrating Factory into Smart Factory Web and example cloud. OUTLOOK The collaboration between KETI and IOSB advances the international usage of standards and architecture patterns for IIoT Project Participants: systems. Benefits are expected for data and service integration and advanced plug Soojin Ji &work, both fundamental for smart KETI factories. Both research organizations conduct training, consultancy and Dr.-Ing. Thomas Usländer development projects for industrial clients. Dr. Ljiljana Stojanovic Information Management and Reference: The two organizations plan to integrate Production Control (ILT) further smart factories and applications Fraunhofer IOSB Karlsruhe into the Smart Factory Web. Sascha Heymann For more information go to: Digital Infrastructure (DIS) www.smartfactoryweb.com. Fraunhofer IOSB-INA Lemgo 1. Federal Ministry for Economic Affairs and Energy (BMWi) and Federal Ministry of Education and Research, Plattform Industrie 4.0: https://www.plattform-i40.de 2. Ministry of Trade, Industry and Energy (MOTIE), The Industrial Innovation Movement 3.0, 29.12.2014 http://english.motie.go.kr 23 vis IT Industrial IoT