Performance Assessment of Hot-Mix Asphalt Modified with Metakaolin as a Partial Filler Replacement: Experimental Testing and Computational Modelling
摘要
The growing need for durable and environmentally responsible pavement materials has driven interest in alternative mineral fillers that can improve the performance of asphalt materials while reducing dependence on conventional resources. This study examines the suitability of metakaolin as a partial substitute for limestone filler in Hot Mix Asphalt and evaluates its influence on mixture volumetric properties, viscoelastic response, stiffness, rutting susceptibility, and surrogate fatigue-related behaviour. Metakaolin was introduced at replacement rates of 0, 5, 10, and 15% by filler mass. The results indicate that metakaolin alters Hot Mix Asphalt response in a stable and frequency-dependent manner. Intermediate metakaolin replacement levels increase stiffness under conditions critical for rutting and markedly enhance resistance to permanent deformation, whereas higher metakaolin contents decrease viscous energy dissipation at intermediate temperatures, suggesting improved surrogate fatigue-related response rather than directly confirmed fatigue-life performance. In parallel, several machine learning approaches were developed to predict dynamic modulus as a function of mixture composition, volumetric parameters, and loading conditions (i.e., temperature and frequency). Thus, ensemble and probabilistic techniques, notably Random Forests and Gaussian Process Regression, provided the strongest predictive performance among the evaluated algorithms within the available dataset and confirmed the dominant roles of loading conditions, with metakaolin influencing the response through its interactions with volumetric characteristics. Collectively, these results demonstrate that metakaolin is a technically viable alternative filler for enhancing Hot Mix Asphalt performance, with potential resource-efficiency benefits that require further environmental validation, and highlight the value of combining performance-based testing with data-driven modelling to support resilient pavement systems.