The successful disputation of our former research assistant Christian Kubik took place on 9 July. His dissertation is entitled ‘Reliable machine learning models for the condition diagnosis of manufacturing processes using the example of blanking.

In his dissertation, Mr. Kubik presents measures to increase the reliability of machine learning models for monitoring the condition of manufacturing processes. The use of data-driven approaches is still inhibited in industry due to various implementation hurdles. Therefore, the full potential of this technology cannot yet be exploited economically. One obstacle is the generalisability of machine learning models to deal with the challenge of changing boundary conditions.

Using the application example of blanking processes, Mr. Kubik investigated how the model error in the data-driven prediction of the wear condition of a shear cutting punch can be minimised. To master a data shift in the life cycle of machine learning models, Mr. Kubik adapts the concepts of robustness, flexibility and resilience and uses various measures to show how these methods can contribute to increasing the reliability of ML-based monitoring systems. With his work, Mr. Kubik shows how future hurdles for the use of ML systems for fault diagnosis in industrial environments can be reduced.

Special thanks go to Prof. Dr. Mathias Liewald from the University of Stuttgart, who supported the doctoral procedure as a co-supervisor.

The PtU would like to thank Mr. Kubik for his many years of commitment, his extraordinary efforts in opening up new research topics and his contribution to the further development of the institute's infrastructure. We wishhim all the best and every success for his professional and private future.