Ostfalia University - Prof. Dr.-Ing. Martin Strube
Ostfalia University - Institute for Mechatronics (IMEC)
Ostfalia University - Institute for Vehicle Engineering Wolfsburg (IFBW)
Lower Saxony Ministry of Science and Culture - Project number: 87140701
The development of components for electric vehicles is giving rise to new requirement profiles, making optimisation increasingly complex. For example, the requirements for the robustness of entire vehicle components have increased due to higher utilisation rates and longer service lives (car sharing, autonomous driving).
At the same time, new materials must be implemented in product design due to sustainability requirements. In addition, cost pressure on the market is very high, meaning that despite efforts to move towards a circular economy, development and manufacturing costs must be reduced significantly.
In technical implementation, it has become apparent that a lack of coupling between CAD design and CAE simulation represents a significant obstacle in the development process. In previous research projects, automated, update-stable CAx process chains were developed, consisting of CAD design and a linked CAE simulation. This link enables automated optimisation of structures.
The special feature of this approach lies in the variation and optimisation in the CAD space, which means that a finished design is available once the optimisation is complete, eliminating the need for post-design of the optimised solution.
Classic algorithms were used for the optimisations, but these have the major disadvantage that the starting conditions influence the optimisation result. Furthermore, the learning process is not utilised in classic optimisation, which means that no empirical knowledge from individual steps can be incorporated into the algorithms used, resulting in long calculation times due to the necessary iterations.
This led to the approach of linking CAx process chains with AI agents in order to implement a learning strategy, thereby making experiential knowledge from previous optimisations usable. This promises the potential to achieve better results in less development time.