Prediction of Performance of Asphalt Overlays Using Decision Tree Algorithm

Pavement performance data are a critical component of pavement management systems, and are affected many factors as the pavement structure, climate, traffic, and material properties. The current traditional parametric techniques still adopted by state agencies can hardly predict performance distress indices reliably and with high accuracy. Hence, the objective of this study was to employ a non-parametric machine learning algorithm to predict the pavement condition index (PCI) of asphalt concrete (AC) overlays on flexible pavement roads over a period of 11 years. In particular, the decision tree (DT) algorithm was examined. A total of 894 log miles from 50 control sections in Louisiana were included in the analysis. Six models were developed using as inputs; the pre-treatment PCI (before overlay application), the highway functional class, the overlay thickness, the overlay age, the cumulative annual truck traffic, the cumulative annual rainfall, and the mean annual temperature. Models yielded a fair accuracy in predicting PCI after 11 years.

Automated summary

This study used a non-parametric machine learning algorithm and multiple variables including pavement structure, climate, and traffic to predict the pavement condition index (PCI) of asphalt concrete overlays on flexible pavement. The decision tree algorithm was examined and achieved fairly accurate predictions of PCI after 11 years.