Datenflussoptimierung mit dynamischer Allokation von Ressourcen in der Industrie
The collection and processing of data, especially streaming data, has a decisive influence on the success of Industry 4.0. They are the basis for new business models and technologies such as condition monitoring or predictive maintenance of industrial plants. To achieve this, it must be ensured that relevant information is available in a timely manner and that the loss of this information is avoided. While the amount of data and the demand for data analysis are increasing, the available resources, especially computing capacity, memory and bandwidth, are still limited.
In this work, essential optimization potentials in streaming data in Industry 4.0 are identified and investigated. In particular, the early extraction of relevant information and the reduction of the amount of data as well as the protection of data are to be mentioned. Specifically, the use of methods of anomaly detection, data compression and distributed ledger technology is examined for this purpose. Furthermore, great potential is seen in optimizing the utilization of limited resources. Due to the decentralized structure of the Industrial Internet of Things, the focus for this is on intelligent multi-agent systems. The evaluation of the individual components proposed in this thesis and their interaction in an overall system shows that the elaborated methods of data processing can be performed closer to the data acquisition due to their high efficiency and thus, information can be provided at an early stage. In addition, the distributed ledger technology IOTA selected in the context of this thesis is evaluated on the basis of two use cases and its applicability on industrial edge devices is confirmed. Furthermore, the optimization of usage of limited resources by multi-agent reinforcement learning systems can significantly enhance the possibility of edge computing in IIoT and prevents data loss due to resource bottlenecks.
The results illustrate that data stream optimization is an important driver for Industry 4.0, as it helps to overcome existing limitations in terms of resources and data safety.