Nondestructive prediction of rutting resistance of in-service middle asphalt layer based on gene expression programing

Abstract

For a multilayered asphalt pavement, rutting resistance of the mixture in the middle asphalt layer underneath the surface course plays a significant role in the high-tempearture stability of the whole pavement structure. Thus, evaluation of the rutting resistance of the middle asphalt layer using field cores is often necessary for project-level pavement maintenance decision-making. However, the extrusion and tests of field cores are time-consuming and destructive to pavement. To address this problem, developing an empirical model to predict the rutting resistance of the middle asphalt layer at a certain service time from historical test results may be a potential alternative. This study aims to address this challenge by using the gene expression programming (GEP) method and a database composed of a large number of multiple-stress repeated load (MSRL) test results of field cores. By correlating the compound creep rate (CCR) of the middle asphalt layer from the MSRL tests to material properties, environmental and traffic parameters, field rutting depth, and middle layer age, the optimal GEP model was developed, and then compared with the conventional multiple linear regression (MLR) model built on the same database. Uncertainty analysis was also conducted through the Monte Carlo simulation (MCs). It was found that the GEP model outperformed the MLR model, with a 6%-7% higher R-square. The uncertainty analysis allows the transport agencies to estimate the reliability of their prediction and make maintenance and rehabilitation (M&R) plans according to the specified reliability target.