ML4P: Machine Learning for Production

Short description of the project

In the lead project Machine Learning for Production (ML4P), we assume that machine learning can be used to optimize performance in modern production facilities - both in the process and piece goods producing industries.

Under the auspices and coordination of the Fraunhofer IOSB, several Fraunhofer Institutes are pooling their application experience and expertise in machine learning to develop solutions for industry. In ML4P, intelligent methods will be formulated to meet the needs of industry and prepare the way for flexible, fast learning systems. A "learning machine" could, for example, involve the installation of intelligent components or efficient, holistic handling of very large volumes of data.

Project goals

• Development of a tool-supported process model
• Realization of a software tool that records and analyzes the current status in order to show possible optimization potentials
• Derivation and selection of suitable methods of machine learning in production

Project results

While the development and integration of the software tools is progressing, the process model has now been published as a short version and is available for free download as a white paper (see below). It describes the way from the problem definition to the continuous operation of the ML-based system comprehensively and includes:

• six phases with clearly defined results,
• the "Machine Learning Pipeline Diagram" and the "virtual process file" as central documents or data structures that represent the current state of knowledge across all phases:
• a role model that includes the disciplines, competencies and functions required in each phase.