Motivation
As a result of Industry 4.0 and the accompanying digitization of production processes, more and more processes in companies have been equipped with sensors and a large amount of data has been collected in recent years. In the process, the digitization process was more technology-driven than benefit-driven. However, the aim of the companies is to use the investments made profitably in order to optimize the overriding goals of costs, time and quality. Despite the increased availability of data, only a fraction of the data has so far been used for process improvement and decision support
Objectives
In the research project “PPM” (Predictive Production Management), the collected data from machines and systems, but also order runs and production planning systems are used to improve production processes on the store floor. The aim of the project is to further develop the digital store floor management software Digital Teamboard from SFM Systems to include the PPM assistance system: a service that automatically detects anomalies in process and machine data as well as KPIs, links these together and makes the resulting findings available to the players in the factory to solve problems.
At the end of the project, the developed PPM service will be integrated into the Digital Teamboard and made marketable. The aim is to achieve commercial exploitation at companies that already use digital store floor management, as well as to build up new customer business. Positive effects for the region will result from new jobs and competitive advantages for participating companies from Hesse.
Approaches
The innovation lies in linking automatic anomaly detection with a subsequent cause-and-effect analysis. Thus, detected anomalies are labeled by workers with appropriate reasons and solution approaches in order to predict future reasons for disturbances based on these labels. Finally, the goal is to automatically detect and localize reasons for malfunctions at an early stage by correlating machine and process data. Both supervised and unsupervised machine learning models are used.
The basis is the “Digital Teamboard” developed by SFM Systems. This collects data at various levels of industrial production and links them together. Key figures for measuring plant productivity, machine data from sensors and explicit domain knowledge of the workers on the store floor can thus be bundled and analyzed. The resulting process and machine expertise can be made available across groups, departments or locations through the analysis.
Acknowledgement
The research project is funded by the Landesoffensive zur Entwicklung Wissenschaftlich-Okonomischer Exzellenz (LOEWE) (HA-Project-No.: 1012/21-14). We thank for the opportunity to work on this project.
Funding source
Consortium partners
Project sponsor