A vehicle body consists of numerous sheet metal components. They undergo several forming processes to achieve the desired shape and function. In production, these forming steps are superimposed by stochastic and non-stationary phenomena, such as material fluctuations and changing process conditions. Engineers refer to these phenomena as process noise, which affects the quality of components. Researchers at the University of Stuttgart are developing new methods to identify process noise as part of the DFG priority program “Data-driven process modelling in metal forming technology” (SPP 2422). By combining process data, expert knowledge, and simulation calculations, it is possible to gain deeper insights into metal forming processes and improve them.
Designing forming technology tools and processes more intelligently
The goal is to make the active surface of forming tools - specifically the surface that comes into direct contact with the material and shape it - more robust and precise. In SPP 2422, existing modeling approaches are therefore being extended by data-driven methods. “The complexity of modern metal forming processes can no longer be fully captured by classical simulation alone,” explains Professor Mathias Liewald, coordinator of the SPP 2422 and head of the Institute for Metal Forming Technology (IFU) at the University of Stuttgart. “We combine real process data with domain-specific knowledge in machine learning models to generate new fundamental insights and make forming processes more robust.”
Triad of forming technology, automation and data science
The SPP 2422 is designed to be interdisciplinary: 13 scientific subprojects across Germany collaborate to model and explain any impact of process noise on metal forming processes. The focus is on metal forming and punching processes commonly used in industry – such as forging, shearing, bending, and deep drawing. All partners collect large amounts of data in their laboratories from simulations of multi-stage process sequences as well as from real endurance tests. These data are analyzed using AI to investigate the impact of process noise, uncover cause-and-effect relationships, and precisely optimize the tool surfaces. Additionally, four working groups foster close interdisciplinary collaboration and focus on cross-cutting tasks within the priority program.
On the way to a holistic assistance system
The overarching research goal is the development of digital and AI-based assistance systems. In the future, these systems will support the design of metal forming tools and processes, laying the foundation for new digital tools and software, known as CAX technologies. The researchers are training new AI models and researching the requirements for data formats, system architectures and interfaces. Over two funding periods until 2029, approaches for improving active tool surfaces are to be developed and transferred into explainable, practical models.
| Contact | Prof. Mathias Liewald MBA, Institute for Metal Forming Technology (IFU), University of Stuttgart |
|---|