The design of sandwich composites with a polyurethane foam core and a metallic face material requires the knowledge of the mechanical properties of the constituent materials. Theses are generally known for metallic materials, but have to be determined for plastic foams, usually via experiments as they are greatly dependent on the foam‘s microstructure. In order to substitute these time-consuming and cost-intensive experiments, this work presents a procedure for characterising the mechanical properties of plastic foams by identifying structure-property linkages using machine learning. The basis for this are experimentally validated simulations of reconstructed and algorithm-based generated digital-twins of polyurethane foam structures. The microstructures of these generated  foam structures are varied systematically to create an information-rich data-basis thereby obtaining an accurate and robust machine-learning tool.

\textbf{Acknowledgements:}
Supported by the ´Helmholtz Artificial Intelligence Cooperation Unit (Helmholtz-AI)’ and funded by the Initiative and Networking Fund (IVF) of the Helmholtz Association (grant number: ZT-I-PF-5-075).