This paper presents a novel deep learning-based methodology for reconstructing the 3D microstructure of small-scale cement samples hydrated in microgravity on the International Space Station (ISS). Using sparse Scanning Electron Microscopy (SEM) data of tri-calcium silicate (C₃S) samples, the method generates a statistically representative ensemble of microstructures. Space-returned samples showed higher porosity (~70%) and elongated portlandite morphology. Reconstructed volumes, validated against micro-CT data, accurately captured the unique microstructural features, providing Representative Volume Elements (RVEs) for mechanical/transport property characterization.

Vishnu Saseendran, Namiko Yamamoto, Peter J. Collins, Aleksandra Radlińska, Sara Mueller, Enrique M. Jackson