Condition monitoring and quality forecasting

Does your production process (still) run smoothly? Or do you (already) have to intervene manually? These questions have become part of daily life for people supervising modern production processes. To answer these questions, current information is required, which is typically collected by sensors directly in the process.

The Department of Machine Vision and Signal Processing implements systems for monitoring production processes with applications in condition monitoring and quality forecasting. The use of these systems makes it possible to detect errors in complex situations at an early stage, as well as to set optimal operating points efficiently.

In practice, the reasons for complexity are manifold: Often, the time required to assess the quality of a component is longer than the actual cycle time, making it difficult to detect errors quickly. In addition, if components can only be inspected using a destructive method, they must be assessed on the basis of random sampling. In order to take full advantage of the available sensor data in such cases, digital signal analysis and machine learning techniques are deployed.

The application spectrum of the monitoring systems is broad, ranging from cyclical productions in plastic injection molding and ultrasonic welding operations to continuous processes in process engineering.

Our experts can also provide support, for example, by selecting and integrating sensors in production as well as in efficiently finding a suitable operating point during commissioning.

Quality forecasting for surface finishes

In plastics processing, there is a growing trend towards applying specific micro- and nanostructured finishes to component surfaces. Full-surface characterization is time-consuming and therefore not an option. However, quality can be forecast after each cycle on the basis of process data, enabling anomalies to be picked up at an early stage.

Prozessüberwachung anhand von Motorspindeldaten

Motorspindeln kommen in der spanenden Fertigung zum Einsatz. Durch integrierte Sensorik können Daten aufgezeichnet werden, die einen Bezug zum aktuellen Zustand der Fertigung (u. a. zum Werkzeug) aufweisen. Mit Methoden des Machine Learnings können kritische Prozesssituationen frühzeitig erkannt werden.

Optimaler Betriebspunkt

Zu Beginn von neuen Fertigungslosen sind in der kunststoffverarbeitenden Industrie häufig Anpassungen der Produktionsparameter erforderlich. Die Gründe liegen beispielsweise in der Zusammensetzung des Rohmaterials, z. B. aufgrund unterschiedlicher Additive oder durch Schwankungen bei der Verwendung von Recyclingmaterial. Das Fraunhofer IPA hat ein System entwickelt, das eine schnelle, zielgerichtete Optimierung des Betriebspunkts ermöglicht.