Mechanical properties of structural foams are largely influenced by the manufacturing process. Depending on the manufacturing process, the foam type (e.g. open-pored/closed-pored), density and the foam orientation are defined. However, these quantities only define the global structural properties. Due to the inhomogeneous process, no adequate statement can yet be made about the real microstructure. However, if a mechanical load is applied, the generated inhomogeneous stress field depends on the actual microstructure, which is characterized by specific values such as pore distribution, mean pore size, spatial gradients in the pore size distribution, pore arrangement and geometry of cell walls. This can lead to local differences in mechanical properties due to irregularities generated by the foaming process.

The objective of the current research is linking the above-mentioned microstructural properties and the resulting macroscopic mechanical material behavior of structural foams by using artificial intelligence (AI) based digital methods. Therefore, polyurethane (PU)-foams of varying densities are analyzed to create a suitable database. For each of the investigated foam densities, the dependency of the microstructure and the mechanical properties according to the sampling position are investigated. First, the mechanical properties of tension and compression load are experimentally determined in parallel and perpendicular spatial direction as well as for varying positions along the foaming directions. The microstructure properties of the real PU-foams are determined using computer tomography scans. Afterwards, the microstructures are reconstructed using computer algorithms developed with the microstructure simulation package Pace3D in order to generate digital twins. Applying various data science methods and micromechanics simulations, the morphological characteristics and the mechanical properties of the digital microstructures are determined. The results from experiments and computational methods are compared and correlations between the mechanical tests and the microstructure analyses are derived. In forthcoming research, this correlation of microstructure and mechanical behavior will be explained and predicted using AI-based methods.