In recent years, the German automotive industry has been experiencing increasingly fierce competition from new market players with low-cost mass products. Press shops must therefore be able to guarantee consistently high component quality in addition to high output. At the same time, forming processes are subject to uncertainty, which can have many different sources. Data-driven approaches make it possible to identify the smallest changes in a process and offer the potential to take appropriate countermeasures. This requires process data in sufficient quantity and high quality.

Previous work has shown that forming processes can be effectively monitored using integrated sensor technology and process data. On this basis, a targeted analysis will now be carried out to determine which sensor signals correlate with process deviations and errors. The aim is to gain a better understanding of signal relevance in order to develop robust models for early fault detection. Various machine learning methods will be applied and compared for this purpose. Modelling and validation will be based on real process data. Both individual signals and various combinations and fusions will be examined as input. In addition, the existing sensor system can be expanded. The exact task and work packages can be agreed upon in a personal meeting!

Research method

Experimental, Theoretisch