Abstract In industrial sheet metal forming processes, the acquisition of accurate data is often challenging due to limited accessibility to critical zones. An example is the correlation of strip position with the resulting force process signal in sheet metal forming, as real production environments offer limited opportunities for direct observation. In progressive forming processes, such as deep-drawing, slight deviations in strip position, caused by errors that occur during feed, can have a significant influence on the quality of the products and thus lead to a differentiation between good and reject. Furthermore, initially occurring errors caused by the prevailing multi-stage process can lead to errors in further processing steps. While force signals in forming processes are easy to acquire, correlating them with the actual process state can be challenging. In contrast to real processes, simulations allow detailed analysis of specific effects and the generation of controlled data sets without disrupting production operations. This paper presents a simulation-driven approach to systematically model common deep-drawing defects in progressive dies due to strip misalignment by varying initial sheet positions in progressive forming simulations, enhancing robustness and enabling defect prediction for process optimisation. Experimental validation ensures that simulation results match real-world behaviour. Using synthetic data sets, the acquired synthetic process data is correlated to real process data with controlled variation of sheet misalignment using machine learning methods. By providing actionable insights into the manufacturing process without disrupting operations, this work lays the foundation for efficient, reliable and adaptive manufacturing systems that advance intelligent forming technologies.
Moske et al. (Mon,) studied this question.