The MöllerGroup, a globally operating family business based in Bielefeld, focuses on injection molding for the automotive industry in its MöllerTech division. To reduce waste and improve sustainability, a Predictive Quality System has been developed to monitor the quality of plastic parts in real time and identify problems early on.
In a pilot project on a selected Engel injection molding system, relevant process parameters were identified and over 600 production orders were evaluated between January 2021 and May 2024. These data included variables such as shift data, machine status, and test variables of the injection processes. Defective parts were also labeled to obtain meaningful training data for machine learning. A particular challenge was the assignment of defective parts to the corresponding production processes, which was made more difficult by time delays.
The analysis focused on the pressure curves and the injection cascade, which can provide information about typical patterns in faulty components. An anomaly detection model was trained to identify deviations in the process data. Initial results showed correlations between certain process variables and waste types, which are now being investigated further.
Future optimizations include a more precise temporal assignment of the data, an extended analysis of the injection cascade with additional modules, and the greater integration of expert knowledge into anomaly detection. Orientation towards Industry 4.0 standards for the systematic recording of quality characteristics is also planned.