Preliminary publication of Chapter 13, published in “KI-Kompass für Entscheider”, Hanser Verlag 2020, ISBN: 978-3-446-46295-3, edited by Ulrich Sendler
Funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany ́s Excellence Strategy – EXC-2023 Internet of Production – 390621612
Authors: Dr.-Ing. Christian Dölle, Stefan Perau, Machine Tool Laboratory WZL of RWTH Aachen University
Due to an increasingly dynamic environment, companies are facing numerous challenges, such as the micro-segmentation of markets and ever shorter product life cycles. One way to address these challenges is to sustainably increase their innovative strength.[1] The embedding of information and communication technologies in physical products, their refinement to so-called cyber-physical systems, offers companies numerous potentials for the realization of innovations. A cyber-physical system enables data-based evaluation of user behavior and the product can be adapted to changing requirements.[2] Such a networking of products is known as the Internet of Things (IoT). The increasing number of connected products offers a potential for manufacturing companies.
Figure 13.1: Concept of the Internet of Production (IoP) (©Werkzeugmaschinenlabor WZL der RWTH Aachen)
In order to realize this potential in manufacturing companies, scientists at RWTH Aachen University have developed the concept of the Internet of Production (IoP) in a cooperation between the fields of production engineering, computer science and economics.[3] The IoP aims to create a new data infrastructure that enables a higher level of cross-domain collaboration by linking all relevant systems along the product life cycle. The concept of the IoP is shown in Figure 13.1.
The IoP is divided horizontally along the product lifecycle into three sub-areas: development, production, and usage. Within the vertical direction of the IoP, the linking and aggregation of company data takes place. The levels “Raw Data”, “Middleware+”, “Smart Data” and “Smart Expert” differ in their respective data granularity. Data granularity means the level of detail or information density of the different levels of the IoP.
The “Raw Data” level, as the lowest level of the IoP, has the finest granularity of data. Proprietary data systems, such as the Product Lifecycle Management (PLM) system, the Enterprise Resource Planning (ERP) system or the Customer Relationship Management (CRM) system, contain the data that represent the system behavior of products and processes along the value chain. However, the existing data records in these systems are mostly indifferent regarding their completeness, accuracy, semantics, and granularity. For cross-system analysis and usage, the data must therefore first be filtered and linked in a targeted manner in an automated process.[4]
Figure 13.2: Concept of digital shadows (©Werkzeugmaschinenlabor WZL der RWTH Aachen)
This task is performed in the IoP at the intermediate level “Middleware+”, where real-time aggregation and synchronization takes place. On the “Smart Data” level, the actual evaluation of the merged data takes place using different algorithms and procedures such as correlation analyses, cluster algorithms, learning algorithms or meta-heuristics. The goal of the “Smart Data” layer is the generation of so-called digital shadows. These represent the relevant correlations along the product life cycle with sufficient accuracy for the respective application. The concept of generating digital shadows and the underlying aggregation of the data is shown in Figure 13.2.
In the digital shadows, an analysis and representation of the system behavior is carried out on the basis of specific questions. In contrast to a Digital Twin, not all system and process details are represented in the Digital Shadow, which leads to a coarser granularity of the data and significantly simplifies and accelerates the processing of the data and representation of information. Both historical analyses and real-time data analyses can be performed to identify optimization potential and verify data-based hypotheses.
Based on the resulting holistic understanding of the system, it is possible to proactively and quickly identify recommendations for action, show the consequences of individual options for action and thus make quick decisions. The processing of the complex interrelationships and representations for decision making are realized on the “Smart Expert” level. Here, the complexity of the information is reduced to required core elements and visually presented to the user.
The development of intuitive user interfaces is crucial for a comfortable use. Procedures of Artificial Intelligence (AI) are provided for on this level explicitly. With the help of AI, application-specific, intelligent agents can be made available, which allow to find more effective and efficient solutions than humans.[5] In addition to the task of decision support, intelligent agents are also able to store decisions already made. These decisions can be set in relation to existing information in order to gain new insights. Thus, the integration of AI creates the possibility for continuous improvement of decision support. Especially a continuous learning of the algorithm based on human decisions and the identification and analysis of recurring problem situations supports a future automated decision making by AI.
The data infrastructure of the IoP ensures to realize the potentials of increasing digitalization for manufacturing companies. Consequently, the IoP provides the necessary infrastructure to integrate applications of artificial intelligence into the value chain of manufacturing companies.
[1] J. Gausemeier, R. Dumitrescu, J. Echterfeld, T. Pfänder, S. Tomas, D. Steffen, F. Thielemann, Innovationen für die Märkte von morgen, Carl Hanser Verlag, München (2019) 3
[2] C. Klötzer, A. Pflaum, Cyber-Physical Systems (CPS) als technologische Basis einer digitalen Supply Chain der Zukunft, In: Becker W. et al. (Hrsg): Geschäftsmodelle in der digitalen Welt, Springer Fachmedien, Wiesbaden (2019) 381-396
[3] G. Schuh, J-P. Prote J-P, S. Dany, Internet of Production, In: G. Schuh et al. (Hrsg): Engineering valley – Internet of production auf dem RWTH Aachen Campus, Festschrift für Univ.-Prof. em. Dr.-Ing. Dipl.-Wirt. Ing. Dr. h.c. mult. Walter Eversheim, 1. Auflage, Apprimus Verlag, Aachen (2017) 1-10
[4] L. Atzori, A. Iera, G. Morabito, The Internet of Things: A Survey, Computer Networks, Vol. 54, No. 15 (2010) 2787-2805
[5] S. J. Russell, P. Norvig, Artificial intelligence: A Modern Approach, Third edition, Pearson Education, Boston, Columbus, Indianapolis (2016) 1-59