Abstract
<title>Abstract</title> <p> This paper presents a statistically rigorous characterisation of pore morphological evolution in reconstituted Kaolinite subjected to one-dimensional consolidation over a wide stress range (100-6,000 kPa). Microstructural behaviour is cross-validated using two independent techniques: scanning electron microscopy (SEM) and mercury intrusion porosimetry (MIP). A coupled SEM-MIP framework is employed to quantify pore descriptors across seven loading stages. While SEM provides two-dimensional projections of pore geometry, MIP offers complementary volumetric measurements with corresponding pore throat diameter, enabling more reliable interpretation through cross-validation. Kaolinite was selected due to its simple mineralogy and well-defined platelet structure, providing a controlled baseline for isolating mechanical loading effects from mineralogical influences. Principal component analysis (PCA) of the full descriptor dataset indicates that three components account for 80.5% of total variance: PC1 (44.8%) is associated with pore size, PC2 (22.0%) with shape, and PC3 (13.7%) with Pore orientation, The PCA space scatter overlap indicating nonlinear evolution with stress and suitability with neural networks training and characterisation. Empirical regression models linking pore descriptors to vertical effective stress, consolidation coefficient (C <sub>v</sub> ), and permeability (k) yield strong correlations (R² = 0.55–0.98). Among these, solidity exhibits the strongest micro-macro coupling, with a logarithmic relationship to C <sub>v</sub> (R² = 0.98) and a power-law relationship to k (R² = 0.97). These results are interpreted within the critical state framework along the oedometric normal compression line. The proposed regression framework provides a quantitative experimental basis for pore-informed constitutive modelling and establishes structured datasets suitable for machine learning applications in clay characterisation. </p>