Today, companies are increasingly challenged by growing dynamism of their environment. This is caused by a higher diversity as well as faster innovation cycles of products and manufacturing systems, by individual customer requirements (lot size of 1), by new legal regulations, or by planned and unexpected disturbances such as the breakdown of machines and employees as well as the interruption of supply chains or the loss of individual sales markets in times of crisis. Due to their complexity, these dynamics can only be adequately dealt with using methods of Artificial Intelligence (AI). These methods enable the early identification of failures of resources (such as machines) through sensor data streams, and furthermore to consider the entirety of the processes executed in a company as a whole and adapt them as required.
In this project, it will be investigated how existing tasks or orders in manufacturing or logistics can be optimized in a goal-oriented way depending on currently available resources and how they can be made resilient to disturbances. For this purpose, resources are continuously monitored in order to determine the current system state and enable prognoses regarding potential failures. The increasing connection to the Internet (Internet of Things) between production plants and their components means that a large number of data streams from sensors and actuators is available to predict a failure of a component and enable countermeasures in the production process. This analysis forms the basis so that this knowledge can be considered in an intelligent process design (planning of processes and maintenance work) and process execution. Humans as actors should be integrated - especially for knowledge-intensive tasks - so that processes in which humans and automated activities are combined are in the center of interest.
In this project, experience-based learning systems are explored and combined with methods of process-oriented case-based reasoning, automatic planning and configuration, and machine learning methods such as deep learning. The application of semantic technologies such as ontologies and industry standards is an important part of this project. Developed research prototypes are tested and demonstrated in realistic application scenarios with the help of a factory simulation plant from Fischertechnik in the IoT Laboratory IoT Laboratory of the University of Trier.
Student Research Assistants
- Florian Brand
- Sascha Stülb
- Arne Kessenich
- Eileen Neumann
- Marcel Mischo, B.Sc.
- Niklas Weingarz, B.Sc.
A list of publications of the EBLS4Industry project can be found here.
A dataset with exemplary failures of the Fischertechnik simulation factory is here available for download. Another process-based dataset with the factory sensor data will be made available soon. In addition, the developed domain ontology (FTOnto) is available here.