DYNAMIC MODULUS PREDICTIVE MODEL BASED ON ARTIFICIAL NEURAL NETWORK FOR THE SUPERPAVE ASPHALT MIXTURES OF NEW MEXICO

In this study, a new dynamic modulus predictive model for the Superpave asphalt-aggregate mixtures of New Mexico is developed based on the artificial neural network methodology. A total of 54 plant-produced asphalt-aggregate mixtures from all over the state were collected, compacted, cored, and sawed to cylindrical test specimens in the laboratory to conduct dynamic modulus testing at different temperatures and loading frequencies. A database containing 1,620 dynamic moduli with phase angles was then used to develop this artificial neural network based predictive model. A neural architecture with 2 twelve-node hidden layers was found to be remarkably suitable for predicting the dynamic modulus and phase angle of asphalt concrete. Statistical evaluation showed that a fairly accurate estimation of dynamic modulus can be attained by using this model.