Big Data in Industrial Printing

This research project investigates what information can be extracted from large amounts of inline sensor data collected by printing presses and to what extent machine learning methods can help in processing and interpreting this data. Machine learning in the context of industrial printing, with its application-specific features, is not yet very widespread. The resulting questions are being addressed in close cooperation with the printing industry.

Specifically, in an interdisciplinary industry research project the dynamic process fluctuations are being investigated together with a paper manufacturer, a printing company and a system manufacturer for web guiding, register control and quality assurance systems using the example of a productive roll-to-roll gravure printing press. Using a dataset (MSDIRPP) of over 40,000 km of printed cardboard produced in this project, application-specific methods are being developed to automatically detect hidden patterns in web run, register quality or winding quality. The knowledge gained from the project should contribute to a better understanding of the process, which can be used, for example, to optimize product quality, reduce waste or predict maintenance.