Quantitative mapping of internal structural parameters to mechanical properties of asphalt concrete using principal component analysis

Characterizing the evolution of internal structural parameters and mechanical properties in asphalt concrete is challenging due to its complex, random internal structure. This study proposes a mapping framework between internal structural "gene units" and mechanical behavior using the material genome concept. Semicircular bending (SCB) tests with Digital Image Correlation (DIC) were performed to extract peak load (Fmax) and maximum horizontal strain (Exx) under controlled conditions. Fourteen internal structure parameters – comprising aggregate contact and pore distribution characteristics – were quantified through digital image processing and 3D X-ray CT reconstruction. Principal component analysis (PCA) followed by multivariate linear regression was employed to correlate structure with performance. The results revealed that the coefficients of variation for macroscopic mechanical indexes and peak strain were 13.14% and 21.01%, respectively. Durbin–Watson statistics of 2.442 (Fmax) and 2.414 (Exx) suggest no significant first-order autocorrelation in the model residuals. The R2 values for the prediction models of peak load and maximum horizontal strain based on internal structural parameters were 0.925 and 0.889. These results demonstrate that the proposed preliminary framework effectively captures the influence of internal structure on asphalt concrete mechanical behaviour, providing a foundation for future data-driven optimisation of mixture design.