DFG-Sonderforschungsbereiche (SFB)
Disciplines
Construction Engineering (30%); Electrical Engineering, Electronics, Information Engineering (20%); Computer Sciences (50%)
Keywords
Lightweight Design,
Fiber-Reincorced Plastics,
Optimization,
Numerical Design
Abstract
This project is a sub-project of a Transregional Collaborative Research Center (CRC)
between Chemnitz University of Technology, RWTH Aachen University, TU Dresden
and TU Wien. The SFB is entitled Intelligent production technologies for lightweight
plastic structures with load-dedicated 3D grading of the reinforcement architecture.
In this CRC, new lightweight construction technologies are to be developed that
allow the continuous transition between different material types, whereby the
material types are particularly concerned with the fiber content in fiber-reinforced
plastics. The key point here is the continuous transition, as the current state of the
art only allows for abrupt transitions, which result in lower component strength. With
the new technology of continuous transitions, components suitable for mass
production are to be produced which, due to their lower weight, have a higher energy
and resource efficiency, especially in mobile applications such as cars or airplanes.
This efficiency should relate to both the manufacturing and use of the product. In the
past, individual solutions have been created to manually demonstrate that such
continuous transitions actually lead to better component properties. This is now to be
generalized and systematized in the CRC. To this end, new numerical design
methods will be developed that predict the optimal fiber distribution, as well as
manufacturing processes that enable the production of these fiber distributions. This
is supplemented by sub-projects dealing with life cycle analysis.
The sub-project funded by the FWF is a project in the field of optimization and
design. We will investigate the extent to which artificial intelligence methods can be
helpful in the design of these components. In particular, the aim is to efficiently
predict component properties and then optimize them in such a way that the
algorithm can learn from past proposed solutions, which is not yet the case.