Bewirb dich bis zum 15. Juli für dein Bachelor- oder Masterstudium an der TU Hamburg.

I3 Junior Project “Merging Computer- and Material Science using Artificial Intelligence”

Audrius Doblies, M. Sc. (former M-11), Janina Mittelhaus M. Sc. (M-11) and Benjamin Boll, M. Sc. (M-EXK2) investigate the mechanical and thermal degradation behavior of small-scaled neat thermosetting matrix material and composites at the institute of polymer composites (M-11, Prof. Bodo Fiedler) and the institute of molecular dynamics simulations of soft matter (M-EXK2, Prof. Robert Meißner).

The motivation for this research is to improve the performance of epoxy based lightweight structures in the renewable energy and transportation sector as one factor to meet global climate goals such as the Paris Agreement. Obviously, a reduction of weight results in fewer emissions. In addition, a better understanding and design of composite structures may become a key enablers for new concepts such as electric propulsion flight or structure integrated batteries with a tremendous reduced impact on emissions. In order to use the superior properties of modern fiber-reinforced polymer (FRP) materials, one has to cope as well with the drawback of the highly complex material. To encounter this, our approach is to study the material behavior by a joined material- and computer science study.
Therefore, several thermosetting matrix materials, like epoxy resins, are exposed to different mechanical loads while the material state is characterized using modern infrared spectroscopy devices. The challenge of highly scattered signals and many influencing factors is overcome by combining specific domain knowledge with relevant machine learning algorithms.
The data acquisition is performed with a Fourier-transformed infrared spectroscope with an appropriate light-sources based on the specimen (MIR and NIR). Afterward, the data is processed and relevant features are extracted. Most intuitive is Gaussian peak fitting to extract the position of the peak in the spectrum as well as the area under the peak. These extracted values can be monitored manually or with machine learning algorithms to study the interaction between bond length, molecular forces and material properties.