Lighthouse project

Fraunhofer-Leitprojekt MED²ICIN
© Elnur Amikishiyevt – stock.adobe.com

ML4P – Machine Learning for Production

In the ML4P lighthouse project, six Fraunhofer Institutes under the coordination of the Fraunhofer Institute for Optronics, System Technologies and Image Exploitation IOSB in Karlsruhe have teamed up to develop a tool-aided process model and to realize relevant interoperable software tools in order to systematically exploit the optimisation potential in production engineering plants through the use of machine learning methods.

Harnessing machine learning (ML) methods, unknown relationships can be learned, processes can be modelled, and adaptive mechanisms can be implemented that make production systems flexible and rapidly modifiable. In contrast to those application domains of machine learning involving enormous volumes of data (image processing, speech recognition, social media, etc.), in the industrial context only “a large quantity of data" is involved, coupled with detailed expert knowledge. With regard to consistent system optimization, both factors must be utilised: all the available data and the full breadth of expertise. In this context not only the trend topic of deep learning is of interest, but also a wide range of other specially adapted ML methods that can work with less data, while simultaneously exploiting previous knowledge.

Machine learning in the production environment

Based on extensive experience in various Fraunhofer Institutes, there is a strong demand for this both in the processing industry and in the piece-goods manufacturing industry – a demand that is often coupled with a lack of ML expertise on the part of users. Accordingly, the Fraunhofer Machine Learning 4 Production lighhouse project aims to specifically address the issues involved in the application of ML methods in the production environment and to develop industrial tools for the efficient use of ML methods. The Fraunhofer Institutes participating in ML4P command excellent ML expertise, an impressive ML method portfolio at a high scientific level, a wealth of application experience in industry as well as know-how in production engineering and materials science.

Tool-aided process model

The research objectives of the Fraunhofer ML4P lighthouse project are to develop a tool-based process model and realise corresponding interoperable software tools in order to systematically capture the relevant knowledge and data of a production facility, identify and evaluate existing optimization potentials, select the most suitable ML methods for specific applications, and apply them to advantage through the use of process data and expertise.

The skills of the Fraunhofer Institutes involved in ML4P complement each other ideally and – in addition to the development of process models, methods and software – allow the project findings to be validated in demanding application domains of industrial production.