--- res: bibo_abstract: - Batch and process fluctuations during the fabrication of sheet metal components result in discrepancies in the resulting component properties, affecting subsequent process steps and potentially leading to production rejects. Consequently, the identification of deviations and knowledge of the effects of fluctuations are crucial for achieving consistently high product quality, reducing waste and thus increasing resource efficiency of production processes through countermeasures derived from this. The approach presented to address this is the use of data-driven metamodeling to map entire process chains and predict process parameters in order to compensate for process and batch fluctuation. The investigated process chain consists of the sub-processes deep drawing, clamping and clinching. For each process step, relevant input and output variables are identified, numerical simulation models are created, and subsequently validated. Variant simulations of the sub-processes are conducted and evaluated to generate a database for the metamodeling of the individual process steps. Machine learning techniques are utilized for the automated selection and optimization of learning methods to create models that depict the relationships between input and output variables. Finally, the models for the sub-processes are linked together to form a superordinate metamodel for the entire process chain, with the aim to make inline-process adaptations possible.