Using Artificial Neural Networks to Predict the Cracking Resistance Change Due to Asphalt Binder Content Variation

Abstract

This paper presents the results of a comprehensive laboratory testing program that was conducted to evaluate the effect of variations in various asphalt mixture properties on changes in the cracking resistance of asphalt mixtures. To this end, a testing factorial consisting of 13 mixes and three asphalt binder content protocols, that is, optimum asphalt binder content (OAC), OAC+0.3%, and OAC−0.3%, were considered. Indirect tension asphalt cracking tests (IDEAL-CT) was conducted on the considered mixtures to examine the interaction effects between mixture characteristics and variations in the binder content on changes in cracking resistance as measured by cracking tolerance index (CTI). Statistical analyses were performed to evaluate the significance of the interaction effects with respect to asphalt mixture type. Regression and artificial neural network models were then developed to predict changes in CTI based on asphalt binder content variation. The results suggested that changes in CTI due to the variation in binder content were significantly influenced by reclaimed asphalt pavement (RAP) content, mixture type, and the type of aggregate used in the production of the asphalt mixture. A good correlation was obtained between mixture characteristics and changes in the CTI values of the asphalt mixtures. The correlation results were further enhanced with the use of artificial neural networks (ANNs).