Project Details
Projekt
Process data-driven modeling for the robustification of shear-cutting collar-pulling processes using effective tool-knit surface design with consideration of edge-crack sensitivity
Subject Area
Primary Shaping and Reshaping Technology, Additive Manufacturing
Data Management, Data-Intensive Systems, Computer Science Methods in Business Informatics
Data Management, Data-Intensive Systems, Computer Science Methods in Business Informatics
Term
since 2023
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 520194818
Collar forming is a well-established forming process in sheet metal processing for creating shaped elements on components that serve to stiffen, distance or fix as well as to join. The pilot hole, which is largely required for this process, is usually made by an upstream shear cutting process. The resulting cutting edge on the components is one of the main causes of edge cracks during subsequent collar forming, in addition to strain hardening in the area of the cutting surface. Despite the extensive knowledge gained from experience and the use of numerical simulation methods in the design of the process chain consisting of shear cutting and collar forming, the often stochastic occurrence of these edge cracks can usually not be predicted and explained. This applies in particular with regard to batch variations of the sheet materials used or to effects of wear on the active tool parts. In the planned research project, therefore, the formation of edge cracks as well as the effective design of the working areas on tools are to be investigated through the interdisciplinary interaction of forming technology and data science by means of a digital representation of the process chain consisting of shear cutting and collar forming. To acquire the necessary database for data-driven process modeling, relevant material, process and machine parameters and the resulting quality characteristic 'width of the edge crack' are recorded in endurance tests in a progressive die along the process chain under consideration and mapped in a corresponding data model. In addition, physical effects and process noise are detected and identified. By analyzing this raw data as well as aggregated data using process analytics and applying machine learning techniques such as deep learning (DL), data quality can be considered and increased at an early stage. The subsequent combination of process data with the domain knowledge (consisting, among others, of expert knowledge as well as results of the FE simulations) in the context of hybrid modeling ultimately leads to the expected transparency, interpretability and explainability of the knowledge-based model and thus of the relationships found between the properties of the semi-finished products, the shear cutting and collar forming parameters as well as the component quality. As a result, a transferable system knowledge is to be created that, taking into account edge crack sensitivity, enables an effective design of the working areas on tools in the process design and thus contributes to a robustification of shear-cutting collar-forming processes.
DFG Programme
Priority Programmes
