Unlocking the potential: analyzing 3D microstructure of small-scale cement samples from space using deep learning.

Area of Science:

• Materials Science
• Civil Engineering
• Aerospace Engineering

Background:

• In-situ resource utilization is key for sustainable extraterrestrial construction.
• Traditional material transport to space is cost-prohibitive.
• Cementitious materials are essential for building structures beyond Earth.

Purpose of the Study:

• To develop a novel methodology for reconstructing cementitious microstructures hydrated in microgravity.
• To utilize deep learning for generating statistically representative microstructures from limited experimental data.
• To enable accurate characterization of materials for space applications.

Main Methods:

• Scanning Electron Microscopy (SEM) was used to capture micrographs of hardened tri-calcium silicate (C3S) samples hydrated in microgravity.
• A deep learning framework was employed to reconstruct 3D microstructural volumes from 2D SEM exemplars.
• Statistical descriptors and micro-CT data were used for validation of reconstructed volumes.

Main Results:

• The deep learning framework successfully generated an ensemble of statistically relevant microstructures.
• Space-hydrated C3S samples exhibited significantly higher porosity (~70%) and elongated portlandite morphology.
• Reconstructed volumes accurately captured the microstructural characteristics of both space-returned and ground-based samples.

Conclusions:

• The developed reconstruction method effectively models microstructures from sparse extraterrestrial data.
• The findings provide a pathway for characterizing mechanical and transport properties of space-based construction materials.
• This research supports the development of advanced materials for long-term human presence in space.