Spatial correlation and pore morphology analysis of limestone calcined clay cement (LC³) via machine learning and image-based characterisation

Comprehending the microstructure of LC³ is of paramount importance since it governs majority properties of cement. Here, we investigate the spatial correlation and pore morphology of LC³, revealing microstructural refinement effects through deep learning and image-based characterisation. A deep learning model was developed to characterise the spatial correlation of the local features of 28-day LC³ with optimised resolution and physical image size, identifying a lower probability of connected pores occurring but a higher likelihood of connected solid particles in LC³ than in OPC. A 33% lower maximum correlation revealed by two-point correlation analysis inferred that LC³ possessed a smaller RVE size and increased packing density. The pore morphological analysis based on BSE images indicated a higher hydration rate and pore deformation in LC³. These findings demonstrate the microstructural refinement mechanisms of LC³ but also lay the foundation for localised microstructural characterisation of cementitious materials with the potential to complement existing traditional analyses.