Process surveillance in hydroforming based on machine learning algorithms

Abstract The requirements related to production and maintenance have increased significantly. An intelligent surveillance of the machines allows the exploitation of significant optimization potentials. Especially hydroforming processes are subject to workpiece specific adjustments and the process parameters have a strong influence on the quality. Furthermore, the surveillance of components which are prone to wear allows the machine operator to optimize maintenance processes in respect of operating costs, machine health and process quality. This paper describes an approach which allows machine operators to use automatically learned data of standard processes and their statistic variances to quickly analyze and optimize hydroforming processes. Since unsupervised machine learning approaches are applied, the algorithm needs only a few process runs to learn the data of standard processes. Furthermore, we introduce algorithms which automatically detect and classify workpiece defects, monitor the wear of hydraulic valves and optimize the energy consumption of the hydraulic pumps.