Machine learning analysis of overweight traffic impact on survival life of asphalt pavement

The objective of this study is to quantify the impact of overweight traffic on asphalt pavement life using machine learning method for survival analysis. Traffic data and field distress measurements were collected from the long-term pavement performance (LTPP) database. A random survival forest algorithm was used to establish predictive models of load-related pavement distresses considering traffic loading, pavement structure, and climate. The variable importance approach was used to select the appropriate variables in the model and reduce prediction error. The findings indicated that the explanatory variables related to axle load spectra and traffic loading were significant in explaining pavement performance degradation. The derived models were further applied to estimate the survival probability curves of asphalt pavement life at different loading scenarios and evaluate the impact of overweight traffic on the reduction of pavement life. The reduction ratio of pavement life due to alligator cracking resulting from overweight traffic was found to be greater than those due to longitudinal cracking and rutting for all the pavement sections. The study findings indicate that the proposed random survival forest model is a promising approach for quantifying the impact of traffic loading on pavement life considering axle load spectra characteristics.