The digital transformation connects the whole value chain with technologies of the Internet of Things (IoT) context. The complexity of software developement which comes with this transformation raises a need for interdisciplinary development of software, hardware and mathematical modeling. Traditional measurement and calibration methods are based on accredited calibration-institutions and standardized procedures. These mechanics have to be adapted to fit the digital transformation.

For this reason the projct FAMOUS combines the expertise of the research institutes with the competence in industrie 4.0 of the companies Endress+Hauser, Lenze and Bosch. Single sensors are conected with a digital twin which is able to to communicate informations of the measurement uncertanty. Sub-networks od sensors are combined with flexible mathematical methods to enable machine-oriented data-evaluation. Organic Computing mehtods enable flexible and autonomous sub-networks.

Contact

In the ADELeS research project, we cooperate with the XITASO GmbH, the Friedrich-Alexander Universität Nürnberg-Erlangen and the Rehau SE funded by the Bavarian Ministry of Economic Affairs, Energy and Technology starting in September 2021. In ADELeS we aim at building an AI-based assistance system to detect and resolve quality defects and other errors within a manufacturing process of extrusion-based machines. This system will strongly interact with the machine operators leading to various requirements on explainability of predictions, suggested actions and the overall usability. Long term, this system will improve overall product quality and mitigate the effects of demographic change and the resulting lack of skilled labor. Therefore, it will futureprove the profitability of the bavarian manufacturing sites.

The envisioned method combines explainable machine learning systems with various forms of expert knowledge and traditional statistical tools. The assistant will interact with operators in the form of passiv suggestions on possible actions as well as by actively adjusting certain aspects of line control directly.

Contact

Faster image processing solution in industry through artificial intelligence

With the High-Tech Strategy 2025 “Research and Innovation for People”, the federal government has set itself the goal of identifying the actual application potential of excellent research more quickly and effectively and making it usable for business and society. For this goal, the bridge between academic research and its economic or social application must be further developed. Thus, the BMBF funded "Validation of the technological and social innovation potential of scientific research - VIP+" supports researchers in systematically validating research results and finding new areas of application. The development of high-precision AI-based image processing solutions - e.g. for the automatic defect detection of components - represents a personnel, time and financial challenge for companies in many branches of industry. Conventional AI methods, such as neural networks, require a lot of time, expert knowledge as well as large data sets to train. Furthermore, such solutions are usually difficult to understand and document, too.

The three project partners in SaMoA have developed a new method that is based on classical image processing solutions and uses nature / genetic inspired algorithms. The method automates large parts of the development process and delivers interpretable solutions with only needing a small amount of data. In addition to saving resources, it also makes later certification of the applications easier, e.g. in aviation. The aim of the project is to validate the method with regard to important performance indicators such as resolution, calculation time, data volume and system limits. Following the project, commercial exploitation is planned via a spin-off/out-licensing and cooperation with industrial partners.

Contact

MOOCS - Metaheuristics for Optimisation of Organic Computing Systems

The application of metaheuristics in OC systems still poses difficulties, especially when trying to find the most suitable metaheuristic or metaheuristics. They should be applicable to all optimisation problems in the OC system and, in addition, self-adaptively respond to changes in the system’s environment.
This project aims improving the application of metaheuristics in OC systems. It will be examined which metaheuristics are suitable for specific problems and, thereby, advance the state of the art through analysing the components of the metaheuristic responsible for these relations. Furthermore, self-adaptive strategies for applying and exchanging these components will be developed, which will lead to a more general optimiser, suitable for many problems common to OC.

Contact