Development of Tree-Based Algorithm for Prediction of Field Performance of Asphalt Concrete Overlays

Abstract

In Louisiana and many other states, the design pavement service life (PSL) for structural asphalt concrete (AC) overlays is typically assumed to be 20 years. However, research studies showed that the actual pavement service life is rarely monitored. This study assessed the field performance of structural overlays in Louisiana and estimated PSL under field conditions. Considering all the distresses, about 75% of the projects had a pavement service life less than 20 years, and the mean PSL for all projects was 17±6 years. This suggests that the overlay thickness designed using a pavement service life of 20 years may not last the entire design period without additional maintenance/rehabilitation activities. Results also showed that random cracking, followed by rutting, were the primary causes of long-term failure in the evaluated projects. A methodology was developed to adjust the designed overlay thickness so that the actual PSL matched the design PSL. This methodology included the development of an extreme gradient boosting (XGB) classification model to predict PSL with an acceptable level of accuracy based on sensitivity, accuracy, and specificity. This approach may be used by transportation agencies in the planning phase of future maintenance and rehabilitation activities and in the design phase of AC overlays. This would also ensure adequate allocation of maintenance and rehabilitation funds by a state agency.