Towards a systematical approach for wear detection in sheet metal forming using machine learning

Abstract Wear is one of the decisive factors for economic efficiency sheet metal forming processes. Thereby, progressive wear phenome lead on hand to a poor workpiece quality and other tool failure resulting in high machine downtimes. This trend intensified by processing high-strength materials reduction lubricant up dry forming. In this context, data-driven monitoring methods such as learning (ML) provide potential detecting at an early stage overcome manual cost-intensive process inspections. The presented study aims ML based inline quantification states within development approach procedure model Knowledge Discovery Time series image data Engineering Epplications (KDT-EA) which validated two processes, blanking roll forming, that strongly differ their physical behavior acquired data. allows estimation with deviation less than 0.83% 2.21% from actual state. Furthermore, it shown combining different feature extraction well compensation unbalanced using augmentation techniques are able improve performance investigated models.