Abstract Cement-based materials often suffer reduced compressive strength due to internal porosity, yet the specific roles of pore size, shape, and spatial distribution remain inadequately quantified. This study investigates the effect of pore geometry and position on the mesoscale compressive strength of a cement-clay mixture. High-resolution X-ray computed tomography (CT) captured three-dimensional pore structure and total porosity in cured specimens. The reconstructed microstructure was used in numerical models in FLAC3D, where a damage-plasticity concrete constitutive model was calibrated against in situ CT-monitored compression tests and validated with a second specimen. A series of parametric simulations were then conducted, systematically varying pore radius (1–2 mm), sphericity (0.79–1.00), and vertical location. The results show that increasing total porosity leads to a pronounced decrease in peak compressive strength (a strong inverse correlation, with exponential trends outperforming linear fits). However, strength in mesoscale should not be predicted by porosity alone: variations in pore size, shape, and position caused significant deviations from the single-parameter trend. To quantify the combined influence of pore-related features, a multiple linear regression model was developed using porosity, pore sphericity and pore position as predictors. The multi-parameter model achieved a high coefficient of determination ( R 2 = 0.80) and a low prediction error (MSE = 0.34 MPa 2 ), representing more than a threefold improvement in accuracy compared to porosity-only correlations. Standardised regression coefficients indicate that porosity is the dominant factor controlling compressive strength, followed by pore position and pore sphericity, all of which are statistically significant. The results demonstrate that pore geometry and spatial distribution critically influence the load-bearing capacity of cementitious mixtures. The proposed integration of CT imaging, physically calibrated numerical modelling, and statistical regression provides a robust framework for quantitative assessment of microstructure–performance relationships at the mesoscale.
Kaczmarczyk et al. (Thu,) studied this question.