How to Merge Process Knowledge and Modern Machine Learning Techniques to Improve Machine Performance

Presented by Nils Barthel, Kampf Schneid- Und Wickeltechnik GmbH & Co. KG

In a time of increased employee turnover and therefore 'deskilling' in many industries, including converters, it is of extreme importance to preserve domain and process knowledge. To maintain high productivity, it is necessary to provide support in the selection of appropriate parameters and make expert knowledge easily accessible for less experienced machine users. One potential solution is to incorporate an assistance system within the machines to automate the procedure. This can be achieved through the specific combination of detailed process knowledge, existing scientific models for describing a winding process and modern data processing techniques. The research fields of "Knowledge Guided Machine Learning" or "Grey Box Modeling" seem promising to achieve the goal of creating more precise models that represent high-speed industry R2R-machinery. This study presents ongoing applied research on how to overcome this problem in the absence of reliable data sources, which is often the case for OEMs that build the machines but no longer have access to their production surroundings.