Prediction of Fracture Toughness of Intermediate Layer of Asphalt Pavements Using Artificial Neural Network

Abstract

For the sustainable management of pavements, predicting the condition of the pavement structure using a consistent and accurate method is necessary. The intermediate layer situated immediately below the surface layer has the greatest effect on the condition of the pavement structure. As a result, to accurately predict the condition of a pavement structure, the mechanistic properties of the intermediate layer are very important. Fracture toughness (FT)—a mechanistic property of the intermediate layer—is an important factor in predicting the distress that develops on the pavement surface. However, measuring FT consistently is practically impossible by coring asphalt pavement over the entire pavement section. Therefore, an artificial neural network (ANN) model—developed by using pavement surface conditions and traffic volume—can predict the FT of the intermediate layer of expressway asphalt pavements. Several ANN models have been developed by applying various optimizers, ANN structures, and preprocessing methods. The optimal model was selected by analyzing the predictive performance, error stability, and distribution of the developed models. The final selected model showed small stable errors, and the distributions of the predicted and measured FTs were similar. Furthermore, FT limits were set with outliers. Future asphalt pavement conditions can be predicted throughout the expressway network by using the proposed model without coring samples.