The generation of extensive labelled data sets for the training of supervised AI methods is very time-consuming and expensive for specific processes. For this reason, data augmentation techniques are used to artificially enlarge data sets and reduce the experimental costs of data acquisition and labelling. Promising results have already been achieved in this context for image data from stamping processes with the help of ‘Generative Adversial Networks (GAN)’. In this thesis, the findings obtained are now to be transferred to time series data from different forming processes. The data sets required for the modelling are already available, so that this is a purely data-based task.

Existing experience in processing data using AI methods is recommended.

Research method

Theoretical