Development of an artificial neural network (ANN)-based model to predict permanent deformation of base course containing reclaimed asphalt pavement (RAP). Issue 11 (2nd November 2021)

• Record Type: Journal Article
• Authors: Ullah, Saad; Tanyu, Burak F.; Zainab, Binte
• Is Part Of: Road materials and pavement design. Volume 22:Issue 11(2021)
• Page Start: 2552
• Page End: 2570
• Publication Date: 2021-11-02
• Subjects: RAP -- unbound base course -- RAP as base course -- sustainable infrastructure -- permanent deformation prediction model -- artificial neural network -- permanent deformation; Road materials -- Periodicals; Highway engineering -- Periodicals; Pavements -- Design and construction -- Periodicals; 625.805
• DOI: 10.1080/14680629.2020.1773304
• Languages: English
• ISSNs: 1468-0629

Abstract

Abstract : Pavement demolition debris is one of the world's major waste problems. Each year the United States produces about 100 million tons of reclaimed asphalt pavement (RAP), out of which more than 60% ends up in landfills or asphalt plants. Recent studies have shown that RAP can be considered a viable alternative to natural base course aggregates to resolve the problem of waste accumulation. In this study efforts have been made to develop an artificial neural network (ANN)-based performance predicting model for base course aggregates blended with RAP. Repeated load triaxial (RLT) tests have been employed in this study to evaluate the performance of base course aggregate. Two different RAP samples were blended with virgin aggregates (VA) in proportions of 20%, 40% and 60% and RLT tests were performed on the RAP-VA blends at three different stress conditions. The data from the laboratory test results were used to model the response of the RAP-VA blends in terms of accumulated permanent deformation against loading cycles. The ANN-based model developed in this study predicted the response of the material with an average coefficient of determination of 0.98. The results indicate that the developed ANN-based model is accurate in comparison to previously published regression models, which do not have the room to accommodate complex material properties as in the case of RAP and other recycled materials.