CHARACTERIZATION OF NON-LINEAR BEHAVIOUR OF ASPHALT CONCRETE AND PREDICTIVE MODELING IN PAKISTAN

Abstract

Proper characterization of asphalt concrete behaviour is important to design pavement, especially in areas with high climate changes such as Pakistan. The conventional techniques of measuring the resilient modulus (M_r) including ASTM D7369-20 take time to complete and do not represent the entire range of non-linear, viscoelastic behaviour. The limitations in this study were solved by creating an Artificial Intelligence (AI) model to forecast the non-linear and temperature-dependent behaviour of asphalt mixtures using local materials obtained in Pakistan. To create a base of information, experiments were performed at 5^oC, 10^oC, 20^oC, and 30^oC. This dataset of 6,082 points was then trained on an Artificial Neural Network (ANN) to predict M_r and cyclic load (P_cyclic) parameter through to 80^oC. The ANN has a high degree of accuracy (target R² above 0.90), which effectively modelled the complicated material response. The main findings can be summarized as the high reduction in M_r with rise in temperature, especially in 10 ^oC to 30 ^oC temperature and the change in state to softened state in high temperatures. This is a hybrid solution that combines both standardized testing and AI as a reliable and effective way to implement sustainable pavement design depending on exclusive environmental and traffic factors in Pakistan.