Forecast of post-overlay flexible pavement roughness progression: a random coefficient linear regression with autocorrelation model

Abstract

Abstract A pavement roughness progression model is an essential component in a pavement management system. Asphalt overlay is a typical pavement maintenance and rehabilitation (M&R) technique. To schedule asphalt overlay in a timely fashion, the relationship between the asphalt overlay design and post-overlay roughness progression is required. However, due to difficulty in data compilation and model development, the effect of endogenous overlay design and continuous variation of asphalt overlay thickness on post-overlay roughness progression is not documented. This study aims to develop a comprehensive post-overlay flexible pavement roughness model with long-term pavement performance (LTPP) data. The asphalt overlay projects from the LTPP SPS-3, SPS-5, and GPS-6 programs were incorporated for data analysis. A random coefficient linear regression with autocorrelation (RCLRA) model is proposed to simultaneously address endogenous overlay design issue, between-section heterogeneity issues, and within-section serial correlation issues in the post-overlay roughness progression. By addressing the within-section serial correlation, the proposed post-overlay roughness model can reduce the mean absolute percentage error (MAPE) from 21.26 percent to 2.19 percent. The model estimation results provide some new insights into the relationship between post-overlay roughness and asphalt overlay design factors. An endogenous overlay design indicator, continuous variation of asphalt overlay thickness, the relative fatigue cracking area, and severe rutting indicator are first identified to have significant effects on post-overlay roughness progression.