Speaker
Description
Punching represents one of the most frequently used manufacturing processes in the sheet metal processing industry. Thereby, one of the most important quality criteria for such punching processes constitutes the geometric shape of the cutting surface. High cutting surface qualities are usually characterized by the lowest possible edge draw in, a high clean cut, as well as a small fracture surface and a low burr height. In this respect, conventional punching processes can only produce clean cut proportions (CCP) up to 20-50% of the sheet thickness. By means of the so called “concave punch nose design” developed at the Institute for Metal Forming Technology, the geometry of conventional punches could be optimized in such a way that the clean-cut proportion along the cutting surface is significantly increased. We will show, that the quality parameters of the sheared component edges thereby show highly nonlinear relationships to the tooling and sheet metal material parameters used. In order to quantify these effects, the data from several numerical punching simulations is used to pretrain an artificial neural network. Input data for the neural network included features of the punch geometry, the size of the clearance, the sheet thickness and the material data of the semi-finished product. The output of the neural network is precise predictions of the achievable cutting surface quality parameters. Our experimentally validated findings show that the process design for novel punching processes is possible, when using a modern machine learning approach.