The first goal of the research group Material Informatics is to develop machine learning methods that can be applied exploratively, i.e. informatics for materials.
- We identify and classify active (smart/intelligent) materials and structures based on their stimulus sensitivity: An input non-mechanical environmental value leads to a mechanical output; the normalized stimulus-strain-curve can be linearized at the working point to extract this sensitivity parameter. The influences from multi-sensitivity in the material can be used to tune this sensitivity in the unified response.
- We find and visualize data for various active materials like Hydrogels, Dielectric Elastomer Actuators, Conductive Polymers, Shape Memory Alloys, Shape Memory Polymers, Piezo-ceramics, Ionic Polymer Metal Composites. The data is applied for the training of different deep learning model classes to mirror their physically described behavior. For data visualization, we experiment with state-of-the-art augmented reality devices.
As a second goal, we focus on the logical behavior of multisensitive active materials, i.e. informatics by materials:
- We use the methods from signal theory to investigate the transfer of information between physical fields. Therefore, we apply continuum-based multi-field models that can accurately describe the interactions, e.g. between the thermal and chemical field. Applications are e.g. an automatically rain-proofing bike-helmet.
- We identify and model multisensitive materials that can act as neurons in an autonomous non-electrical deep learning network.
If you are interested in our group's activities, please contact us!