Predicting fine blanking process signals from sheet metal thickness

Abstract

Abstract. In sheet metal forming and blanking processes, the direct assessment of process conditions and product quality poses a challenge due to high production rates and the inaccessibility of the tool. In this context, the process signals generated by the manufacturing process, such as force and acoustic emissions, have the potential to serve as a valuable source of information, containing important insights into the quality of the final product as well as the complexity of the process itself. To date, it is not yet fully understood how these process signals depend on different influencing factors, such as process parameters. However, knowing how process signals, which reflect the process state, change with influencing factors is relevant to put observed signals into context and make informed decisions with respect to parameter adjustments. Conditional generative AI models, such as conditional generative adversarial networks (CGANs) offer a promising approach to the aforementioned issue by generating probable process signals based on specified conditions. In this study, thin metal sheets with three different thicknesses were provided into a fine blanking process, and corresponding punching force signals were measured. With these signals, a conditional-deep convolutional GAN (C-DCGAN), a model that combines the principles of both CGAN and deep convolutional GAN (DCGAN), is trained with sheet metal thickness specified as a condition. The trained generator is employed to predict process signals for different sheet thickness values. The presented model is evaluated with respect to thickness values that were known during training time as well as with thickness values that were not presented to the model during training.